Analyzing the External Environment for e-Business Andreas Bitros
This document proposes a systemic approach for analyzing a company's external environment for e-business strategies. It discusses limitations of traditional PEST and Porter's Five Forces analyses and introduces the concept of "strategic points" to map interactions between key external factors. The approach is demonstrated through case studies of Orbitz and Netflix. While the strategic points concept helps treat different groups as having multiple roles, the model could be improved by further defining connection types between points and adding regulatory/user perspectives over time. More case studies are needed to generalize the findings and develop the approach further.
Drivers of e business value creation inIJMIT JOURNAL
With the development and growth of internet, its applications of e-banking, e-commerce, and e-business
became irreplaceable channels regarding its fast access, rich content, and smooth interactivity. High
investments are paid toward improving the quality of service offered by the banks. This paper is dedicated
to empirically investigating the drivers of e-Business value creation in the Jordanian banking sector. This
work summarizes the main differences among employees of Jordanian and foreign bank regarding their
perspectives. Many of the competing foreign banks to the Jordanian banks are enforced with huge financial
capital, having long periods of banking practices and are employing cutting-edge technologies and tools.
To minimize the technological gap, Jordanian banks are working hard to develop their e-Business services.
This in one hand has to enhance their trust, satisfaction, and commitment toward existing customers and
entice new comers on other hand. Based on business model of Amit and Zott, i.e. the four constructs of e-
Value framework (efficiency, complementarities, lock-in, and novelty), four hypotheses have been
formulated to test the differences in the drivers of e-Business value creation between Jordanian and foreign
banks. A survey questionnaire in a form of paper-and-pencil was delivered personally to 200 employees
from four main Jordanian banks and 200 employees from four foreign banks working in Jordan. The
questionnaire was formed and constructed to test the proposed hypotheses. the findings in this study based
on the SEM and T-test analyses, revealed important implications that will help banks’ managers to make
well-informed decisions and policies regarding investments and resources allocation for implementing e-
Business strategies and ventures. The paper concludes with discussing the importance of these findings for
practitioners and for future research on value accrued from e-Business services.
This document describes a fuzzy logic-based decision support system to help business executives select appropriate e-business models. It discusses existing frameworks for classifying e-business models and their limitations. The proposed system uses fuzzy logic to capture executives' linguistic assessments of key business measures. It assesses each business dimension using fuzzy calculations and rules. A prototype was tested with positive feedback from business executives. The system aims to assist decision-making by narrowing the search for suitable e-business models.
Effects of unionization on workplace safety enforcementTaxScout Inc.
We study how unionization affects the enforcement of workplace safety laws. To generate credible causal estimates, we use a regression discontinuity design comparing
outcomes in establishments where unions just won representation elections to outcomes in establishments where union just lost such elections.
This document discusses the rise of predictive analytics and its value in enterprise decision making. It begins by explaining how predictive analytics has expanded from niche uses to a widely adopted competitive technique, fueled by big data, improved analytics tools, and demonstrated successes. A classic example given is credit scoring, which uses predictive models to assess credit risk. The document then provides examples of other areas where predictive models generate value, such as marketing, customer retention, pricing, and fraud prevention. It discusses how effective predictive models are built by using statistical techniques on data that describes predictive factors and outcomes. The document argues that predictive models provide the most value when applied to processes involving large volumes of similar decisions that have significant financial or other impacts, and where relevant electronic
Analyzing the External Environment for e-Business Andreas Bitros
This document proposes a systemic approach for analyzing a company's external environment for e-business strategies. It discusses limitations of traditional PEST and Porter's Five Forces analyses and introduces the concept of "strategic points" to map interactions between key external factors. The approach is demonstrated through case studies of Orbitz and Netflix. While the strategic points concept helps treat different groups as having multiple roles, the model could be improved by further defining connection types between points and adding regulatory/user perspectives over time. More case studies are needed to generalize the findings and develop the approach further.
Drivers of e business value creation inIJMIT JOURNAL
With the development and growth of internet, its applications of e-banking, e-commerce, and e-business
became irreplaceable channels regarding its fast access, rich content, and smooth interactivity. High
investments are paid toward improving the quality of service offered by the banks. This paper is dedicated
to empirically investigating the drivers of e-Business value creation in the Jordanian banking sector. This
work summarizes the main differences among employees of Jordanian and foreign bank regarding their
perspectives. Many of the competing foreign banks to the Jordanian banks are enforced with huge financial
capital, having long periods of banking practices and are employing cutting-edge technologies and tools.
To minimize the technological gap, Jordanian banks are working hard to develop their e-Business services.
This in one hand has to enhance their trust, satisfaction, and commitment toward existing customers and
entice new comers on other hand. Based on business model of Amit and Zott, i.e. the four constructs of e-
Value framework (efficiency, complementarities, lock-in, and novelty), four hypotheses have been
formulated to test the differences in the drivers of e-Business value creation between Jordanian and foreign
banks. A survey questionnaire in a form of paper-and-pencil was delivered personally to 200 employees
from four main Jordanian banks and 200 employees from four foreign banks working in Jordan. The
questionnaire was formed and constructed to test the proposed hypotheses. the findings in this study based
on the SEM and T-test analyses, revealed important implications that will help banks’ managers to make
well-informed decisions and policies regarding investments and resources allocation for implementing e-
Business strategies and ventures. The paper concludes with discussing the importance of these findings for
practitioners and for future research on value accrued from e-Business services.
This document describes a fuzzy logic-based decision support system to help business executives select appropriate e-business models. It discusses existing frameworks for classifying e-business models and their limitations. The proposed system uses fuzzy logic to capture executives' linguistic assessments of key business measures. It assesses each business dimension using fuzzy calculations and rules. A prototype was tested with positive feedback from business executives. The system aims to assist decision-making by narrowing the search for suitable e-business models.
Effects of unionization on workplace safety enforcementTaxScout Inc.
We study how unionization affects the enforcement of workplace safety laws. To generate credible causal estimates, we use a regression discontinuity design comparing
outcomes in establishments where unions just won representation elections to outcomes in establishments where union just lost such elections.
This document discusses the rise of predictive analytics and its value in enterprise decision making. It begins by explaining how predictive analytics has expanded from niche uses to a widely adopted competitive technique, fueled by big data, improved analytics tools, and demonstrated successes. A classic example given is credit scoring, which uses predictive models to assess credit risk. The document then provides examples of other areas where predictive models generate value, such as marketing, customer retention, pricing, and fraud prevention. It discusses how effective predictive models are built by using statistical techniques on data that describes predictive factors and outcomes. The document argues that predictive models provide the most value when applied to processes involving large volumes of similar decisions that have significant financial or other impacts, and where relevant electronic
Case Study, Cairo-Amman Bank-Jordan: Improving an Organization by the use of ...journal ijrtem
Abstract : In this study, analysis and comparison between the most important enterprise systems Enterprise Resource Planning (ERP), Supply Chain Management (SCM), and Customer Relationship Management (CRM) are studied in accordance with the organization innovation performance as a special case Cairo Amman Bank in Jordan for process and as well the CRM while using smart phones and iPads as product innovations. Moreover, this study employed a database obtained from two computers aided, mobile surveys conducted since 2010 by the Cairo Amman Bank in Jordan. Both studies are focused along the dispersal and use of the information and communication technology (ICT) in Cairo Amman Bank and its ramifications. Each database contained information about 4000 customers with more than five employees, these databases representatively choose from the most important services in the bank. These databases are taken for the sample which is furnished from the credit rating employees or agents. This agency who provides the biggest database on organizations available in Jordan. Moreover, collects basic information about customers as addresses, sectors, organization’s sizes in the market, and sizes in all enterprises that applied to the bank for credits. These excerpts from the populations of Jordan organizations were stratified according to two size categories, as Jordan and Palestine (East/West) and to several branches in both states. As many organizations as needed have been asked until all classes were met. The interviewee was, in general, the principal executive officer of the organizations who could also resolve to legislate on queries to a corresponding employee like, e. g.,
This study examines the relationship between goodwill impairment losses and bond credit ratings using regression analysis. The results show a negative relationship, such that firms with higher goodwill impairment losses receive lower credit ratings. Additional tests considering market conditions, changes over time, and addressing endogeneity support this finding. This study contributes to the literature on goodwill impairment and bond credit ratings by being the first to directly examine the relationship between the two at the firm level. The findings suggest that credit rating agencies may use goodwill impairment information when assessing firm creditworthiness.
SEM Statistical Analyses in Management Strategiesijdms
This paper was to analyse what factors will affect online shoppers’ purchase intention in the e-business environment. This paper discussed a number of conceptual models and following hypotheses. Two typical e-business companies (Alibaba and Amazon) were ideal comparative analysis models. Structural Equation Modelling (SEM) and the factor analysis were adopted for statistical and empirical analyses. The results showed highly positive correlations among the identified factors. The factors indicated a large innovative performance influence. The factors also verified the huge impact in the sustain ability development. These results may provide some vision on how e-business companies can win a competition and increase their
profit
This white paper introduces the Industry Building Blocks (IBB) classification system, which divides the global economy into over 17,000 ultra-granular industries. IBB classifications are based on Michael Porter's five competitive forces and define industries based on unique combinations of competitors, buyers, suppliers, substitutes, and potential entrants. This creates a more granular system than traditional industry classifications. The paper argues that understanding industries at this level of granularity provides insights into companies, themes, and mergers that are not possible with less granular classifications. It illustrates how IBB can be used to analyze companies like Apple, which it divides into 50 distinct industries based on competitive dynamics.
Types of Information Technology Capabilities and Their Role in Competitive Ad...Yung-Chun Chang
The document summarizes a study that examined the relationship between different types of information technology (IT) capabilities and competitive advantage. It distinguishes between value capabilities like IT infrastructure, competitive capabilities like IT business experience and relationship infrastructure, and dynamic capabilities like organizational learning. The study found that while IT infrastructure did not relate to competitive advantage, IT business experience and relationship infrastructure did positively impact competitive advantage. Organizational learning also positively impacted the development of other IT capabilities. The results suggest that certain IT management capabilities can provide firms a competitive edge.
Clustering Prediction Techniques in Defining and Predicting Customers Defecti...IJECEIAES
With the growth of the e-commerce sector, customers have more choices, a fact which encourages them to divide their purchases amongst several ecommerce sites and compare their competitors‟ products, yet this increases high risks of churning. A review of the literature on customer churning models reveals that no prior research had considered both partial and total defection in non-contractual online environments. Instead, they focused either on a total or partial defect. This study proposes a customer churn prediction model in an e-commerce context, wherein a clustering phase is based on the integration of the k-means method and the Length-RecencyFrequency-Monetary (LRFM) model. This phase is employed to define churn followed by a multi-class prediction phase based on three classification techniques: Simple decision tree, Artificial neural networks and Decision tree ensemble, in which the dependent variable classifies a particular customer into a customer continuing loyal buying patterns (Non-churned), a partial defector (Partially-churned), and a total defector (Totally-churned). Macroaveraging measures including average accuracy, macro-average of Precision, Recall, and F-1 are used to evaluate classifiers‟ performance on 10-fold cross validation. Using real data from an online store, the results show the efficiency of decision tree ensemble model over the other models in identifying both future partial and total defection.
This document discusses the role and skills needed for Six Sigma Black Belts. It provides examples of projects Black Belts have led, including analyzing website download times, understanding customer retention, and reducing equipment delivery cycle times. The document aims to define what Black Belts do and the technical skills required. It recommends a curriculum for Black Belts and compares it to requirements for Certified Quality Engineers and a master's in applied statistics. Other issues around developing Black Belts like candidate selection and mentoring are also discussed.
This chapter discusses tools for assessing a firm's internal environment, including SWOT analysis, value chain analysis, resource-based view, and financial analysis. It describes how SWOT has limitations and value chain analysis examines primary and support activities. A firm's resources must be valuable, rare, inimitable, and non-substitutable for competitive advantage. Financial ratios, balanced scorecards, and stakeholder perspectives provide methods to evaluate firm performance.
Yes Virginia! A Profile In Excellence White PaperJon Hansen
Who Can Benefit from this Paper?
This thought provoking white paper is an essential resource tool for public sector organizations that are already in the midst of an established program, or ones who are contemplating a change. Although they do not operate within the same framework of a public or government entity, private sector companies can also gain important insight as the paper’s principles are universal in their applicability.
Utilizing an advanced research methodology, the primary objective of this paper is to provide policy-makers (and those affected by government policy) with a multi-dimensional “objective lens” through which they will be able to view the veracity of both existing as well as contemplated initiatives. The resulting insights will empower program champions to take the necessary steps to deliver tangible and sustainable results.
ifib Lunchbag: CHI2018 Highlights - Algorithms in (Social) Practice and morehen_drik
The document summarizes several papers presented at CHI 2018 on the topics of:
1) Understanding user experience of co-creation with AI through a drawing collaboration study.
2) Perceptions of justice and fairness in algorithmic decision-making through experimental studies.
3) A qualitative study of perceptions of algorithmic fairness among marginalized groups.
4) The effects of communicating advertising algorithm processes on user perceptions and trust.
This document summarizes research on the adoption of e-procurement in Hong Kong. It begins by defining e-procurement and describing how it is changing procurement practices. It then reviews literature on the benefits of e-procurement such as reduced costs and increased efficiency. The document presents results from the authors' empirical study on the critical success factors and barriers to adopting e-procurement in Hong Kong. It finds that educating companies on both long and short-term benefits can encourage adoption. Critical success factors include financial support, compatibility with existing systems, management support, understanding company priorities, and security systems. The framework developed from this study can help other countries implement successful e-procurement policies and strategies.
This document discusses research conducted by SystemicLogic on implementing product line concepts in large banks. The researchers found that while product line practices developed by SEI/CMU were relevant, a move toward "service lines" was needed to address banks' delivery of financial services through software. Key challenges identified in applying product line approaches included banks' huge invested infrastructures organized around traditional product silos, complex legacy systems, and skeptical perceptions of theoretical concepts. The researchers concluded the term "product" has different meanings, and a service line concept aligned better with banks' delivery of financial services through continually updated software processes and channels rather than discrete software products.
Business Talk: Harnessing Generative AI with Data Analytics MaturityIJCI JOURNAL
Generative AI applications offer transformative potential for business operations, yet their adoption introduces substantial challenges. This paper utilizes the CBDAS data maturity model to pinpoint pivotal success factors for seamless generative AI integration in businesses. Through a comprehensive analysis of these factors, we underscore the essentials of generative AI deployment: cohesive architecture, robust data governance, and a data-centric corporate ethos. The study also highlights the hurdles and facilitators influencing its implementation. Key findings suggest that fostering a data-friendly culture, combined with structured governance, optimizes generative AI adoption. The paper culminates in presenting the practical implications of these insights, urging further exploration into the real-world efficacy of the proposed recommendations.
The document discusses standards that must be followed by Wright Aircraft Corp to enable an effective information security program, noting that compliance is mandatory though deviation is possible with approval. The standards define minimum baseline procedures, practices, and configurations for systems and related topics to provide a single reference point during various stages of development and contracting. However, the standards do not provide detailed instructions for how to meet the company's policies.
When businesses integrate artificial intelligence (AI) solutions, they begin a transformative journey that promises to automate operations, enhance decision-making, and personalize customer experiences. Yet, deploying AI is only the beginning. The true challenge—and opportunity—lies in continuously evaluating and optimizing these solutions to ensure they deliver maximum value and remain aligned with evolving business goals and market conditions.
This document summarizes an IBM executive report on using business analytics to gain competitive advantage. It discusses how analytics can help organizations understand customer behavior, risks, and regulations to inform strategic decision making. The report finds that while technology barriers to analytics are decreasing, organizational culture challenges remain, such as integrating data across departments and establishing a leadership mandate to make decisions based on facts. It recommends that organizations lay the foundation for fast responses, extract value by aligning objectives with integrated data, and use predictive analytics to detect opportunities. When applied to understanding customers, risks, and regulations, analytics can help optimize performance in today's complex environment.
Original definition Predictive Analytics SPSS Jan 15, 2003 Intriduction SlidesJaap Vink
In 2003 the SPSS Senior Management & Marketing under the leadership of Jack Noonan, Dyke Hensen & Matt Cutler coined the phrase "Predictive Analytics" to explain to the market and to the analysts how SPSS differed from BI companies like BO and Cognos. This file contains the presentation by Matt Cutler introducing PA and the definition to the SPSS employees on January 15 2003
This document summarizes an ethics and governance assessment of JP Morgan Chase conducted by EthicsGrade. In Q2 2022, JP Morgan Chase scored 63.57 out of 100 on the EthicsGrade assessment, placing them in the 83rd percentile of all rated companies and 3rd among their peers. The assessment evaluated 344 data points across categories like structure, public policy, data privacy, and sustainability.
Application of AI in customer relationship managementShashwat Shankar
The document provides an overview of applying artificial intelligence (AI) techniques in customer relationship management (CRM). It first discusses the importance and goals of CRM, including understanding customer behavior to improve acquisition, retention, loyalty and profitability. It then classifies CRM into operational and analytical approaches. The document proposes a classification framework for AI techniques in CRM, including customer identification, attraction, retention and development. It describes common AI models like association, classification, clustering, forecasting and regression. Finally, it discusses using this framework to systematically review how AI can make CRM more effective.
Case Study, Cairo-Amman Bank-Jordan: Improving an Organization by the use of ...journal ijrtem
Abstract : In this study, analysis and comparison between the most important enterprise systems Enterprise Resource Planning (ERP), Supply Chain Management (SCM), and Customer Relationship Management (CRM) are studied in accordance with the organization innovation performance as a special case Cairo Amman Bank in Jordan for process and as well the CRM while using smart phones and iPads as product innovations. Moreover, this study employed a database obtained from two computers aided, mobile surveys conducted since 2010 by the Cairo Amman Bank in Jordan. Both studies are focused along the dispersal and use of the information and communication technology (ICT) in Cairo Amman Bank and its ramifications. Each database contained information about 4000 customers with more than five employees, these databases representatively choose from the most important services in the bank. These databases are taken for the sample which is furnished from the credit rating employees or agents. This agency who provides the biggest database on organizations available in Jordan. Moreover, collects basic information about customers as addresses, sectors, organization’s sizes in the market, and sizes in all enterprises that applied to the bank for credits. These excerpts from the populations of Jordan organizations were stratified according to two size categories, as Jordan and Palestine (East/West) and to several branches in both states. As many organizations as needed have been asked until all classes were met. The interviewee was, in general, the principal executive officer of the organizations who could also resolve to legislate on queries to a corresponding employee like, e. g.,
This study examines the relationship between goodwill impairment losses and bond credit ratings using regression analysis. The results show a negative relationship, such that firms with higher goodwill impairment losses receive lower credit ratings. Additional tests considering market conditions, changes over time, and addressing endogeneity support this finding. This study contributes to the literature on goodwill impairment and bond credit ratings by being the first to directly examine the relationship between the two at the firm level. The findings suggest that credit rating agencies may use goodwill impairment information when assessing firm creditworthiness.
SEM Statistical Analyses in Management Strategiesijdms
This paper was to analyse what factors will affect online shoppers’ purchase intention in the e-business environment. This paper discussed a number of conceptual models and following hypotheses. Two typical e-business companies (Alibaba and Amazon) were ideal comparative analysis models. Structural Equation Modelling (SEM) and the factor analysis were adopted for statistical and empirical analyses. The results showed highly positive correlations among the identified factors. The factors indicated a large innovative performance influence. The factors also verified the huge impact in the sustain ability development. These results may provide some vision on how e-business companies can win a competition and increase their
profit
This white paper introduces the Industry Building Blocks (IBB) classification system, which divides the global economy into over 17,000 ultra-granular industries. IBB classifications are based on Michael Porter's five competitive forces and define industries based on unique combinations of competitors, buyers, suppliers, substitutes, and potential entrants. This creates a more granular system than traditional industry classifications. The paper argues that understanding industries at this level of granularity provides insights into companies, themes, and mergers that are not possible with less granular classifications. It illustrates how IBB can be used to analyze companies like Apple, which it divides into 50 distinct industries based on competitive dynamics.
Types of Information Technology Capabilities and Their Role in Competitive Ad...Yung-Chun Chang
The document summarizes a study that examined the relationship between different types of information technology (IT) capabilities and competitive advantage. It distinguishes between value capabilities like IT infrastructure, competitive capabilities like IT business experience and relationship infrastructure, and dynamic capabilities like organizational learning. The study found that while IT infrastructure did not relate to competitive advantage, IT business experience and relationship infrastructure did positively impact competitive advantage. Organizational learning also positively impacted the development of other IT capabilities. The results suggest that certain IT management capabilities can provide firms a competitive edge.
Clustering Prediction Techniques in Defining and Predicting Customers Defecti...IJECEIAES
With the growth of the e-commerce sector, customers have more choices, a fact which encourages them to divide their purchases amongst several ecommerce sites and compare their competitors‟ products, yet this increases high risks of churning. A review of the literature on customer churning models reveals that no prior research had considered both partial and total defection in non-contractual online environments. Instead, they focused either on a total or partial defect. This study proposes a customer churn prediction model in an e-commerce context, wherein a clustering phase is based on the integration of the k-means method and the Length-RecencyFrequency-Monetary (LRFM) model. This phase is employed to define churn followed by a multi-class prediction phase based on three classification techniques: Simple decision tree, Artificial neural networks and Decision tree ensemble, in which the dependent variable classifies a particular customer into a customer continuing loyal buying patterns (Non-churned), a partial defector (Partially-churned), and a total defector (Totally-churned). Macroaveraging measures including average accuracy, macro-average of Precision, Recall, and F-1 are used to evaluate classifiers‟ performance on 10-fold cross validation. Using real data from an online store, the results show the efficiency of decision tree ensemble model over the other models in identifying both future partial and total defection.
This document discusses the role and skills needed for Six Sigma Black Belts. It provides examples of projects Black Belts have led, including analyzing website download times, understanding customer retention, and reducing equipment delivery cycle times. The document aims to define what Black Belts do and the technical skills required. It recommends a curriculum for Black Belts and compares it to requirements for Certified Quality Engineers and a master's in applied statistics. Other issues around developing Black Belts like candidate selection and mentoring are also discussed.
This chapter discusses tools for assessing a firm's internal environment, including SWOT analysis, value chain analysis, resource-based view, and financial analysis. It describes how SWOT has limitations and value chain analysis examines primary and support activities. A firm's resources must be valuable, rare, inimitable, and non-substitutable for competitive advantage. Financial ratios, balanced scorecards, and stakeholder perspectives provide methods to evaluate firm performance.
Yes Virginia! A Profile In Excellence White PaperJon Hansen
Who Can Benefit from this Paper?
This thought provoking white paper is an essential resource tool for public sector organizations that are already in the midst of an established program, or ones who are contemplating a change. Although they do not operate within the same framework of a public or government entity, private sector companies can also gain important insight as the paper’s principles are universal in their applicability.
Utilizing an advanced research methodology, the primary objective of this paper is to provide policy-makers (and those affected by government policy) with a multi-dimensional “objective lens” through which they will be able to view the veracity of both existing as well as contemplated initiatives. The resulting insights will empower program champions to take the necessary steps to deliver tangible and sustainable results.
ifib Lunchbag: CHI2018 Highlights - Algorithms in (Social) Practice and morehen_drik
The document summarizes several papers presented at CHI 2018 on the topics of:
1) Understanding user experience of co-creation with AI through a drawing collaboration study.
2) Perceptions of justice and fairness in algorithmic decision-making through experimental studies.
3) A qualitative study of perceptions of algorithmic fairness among marginalized groups.
4) The effects of communicating advertising algorithm processes on user perceptions and trust.
This document summarizes research on the adoption of e-procurement in Hong Kong. It begins by defining e-procurement and describing how it is changing procurement practices. It then reviews literature on the benefits of e-procurement such as reduced costs and increased efficiency. The document presents results from the authors' empirical study on the critical success factors and barriers to adopting e-procurement in Hong Kong. It finds that educating companies on both long and short-term benefits can encourage adoption. Critical success factors include financial support, compatibility with existing systems, management support, understanding company priorities, and security systems. The framework developed from this study can help other countries implement successful e-procurement policies and strategies.
This document discusses research conducted by SystemicLogic on implementing product line concepts in large banks. The researchers found that while product line practices developed by SEI/CMU were relevant, a move toward "service lines" was needed to address banks' delivery of financial services through software. Key challenges identified in applying product line approaches included banks' huge invested infrastructures organized around traditional product silos, complex legacy systems, and skeptical perceptions of theoretical concepts. The researchers concluded the term "product" has different meanings, and a service line concept aligned better with banks' delivery of financial services through continually updated software processes and channels rather than discrete software products.
Business Talk: Harnessing Generative AI with Data Analytics MaturityIJCI JOURNAL
Generative AI applications offer transformative potential for business operations, yet their adoption introduces substantial challenges. This paper utilizes the CBDAS data maturity model to pinpoint pivotal success factors for seamless generative AI integration in businesses. Through a comprehensive analysis of these factors, we underscore the essentials of generative AI deployment: cohesive architecture, robust data governance, and a data-centric corporate ethos. The study also highlights the hurdles and facilitators influencing its implementation. Key findings suggest that fostering a data-friendly culture, combined with structured governance, optimizes generative AI adoption. The paper culminates in presenting the practical implications of these insights, urging further exploration into the real-world efficacy of the proposed recommendations.
The document discusses standards that must be followed by Wright Aircraft Corp to enable an effective information security program, noting that compliance is mandatory though deviation is possible with approval. The standards define minimum baseline procedures, practices, and configurations for systems and related topics to provide a single reference point during various stages of development and contracting. However, the standards do not provide detailed instructions for how to meet the company's policies.
When businesses integrate artificial intelligence (AI) solutions, they begin a transformative journey that promises to automate operations, enhance decision-making, and personalize customer experiences. Yet, deploying AI is only the beginning. The true challenge—and opportunity—lies in continuously evaluating and optimizing these solutions to ensure they deliver maximum value and remain aligned with evolving business goals and market conditions.
This document summarizes an IBM executive report on using business analytics to gain competitive advantage. It discusses how analytics can help organizations understand customer behavior, risks, and regulations to inform strategic decision making. The report finds that while technology barriers to analytics are decreasing, organizational culture challenges remain, such as integrating data across departments and establishing a leadership mandate to make decisions based on facts. It recommends that organizations lay the foundation for fast responses, extract value by aligning objectives with integrated data, and use predictive analytics to detect opportunities. When applied to understanding customers, risks, and regulations, analytics can help optimize performance in today's complex environment.
Original definition Predictive Analytics SPSS Jan 15, 2003 Intriduction SlidesJaap Vink
In 2003 the SPSS Senior Management & Marketing under the leadership of Jack Noonan, Dyke Hensen & Matt Cutler coined the phrase "Predictive Analytics" to explain to the market and to the analysts how SPSS differed from BI companies like BO and Cognos. This file contains the presentation by Matt Cutler introducing PA and the definition to the SPSS employees on January 15 2003
This document summarizes an ethics and governance assessment of JP Morgan Chase conducted by EthicsGrade. In Q2 2022, JP Morgan Chase scored 63.57 out of 100 on the EthicsGrade assessment, placing them in the 83rd percentile of all rated companies and 3rd among their peers. The assessment evaluated 344 data points across categories like structure, public policy, data privacy, and sustainability.
Application of AI in customer relationship managementShashwat Shankar
The document provides an overview of applying artificial intelligence (AI) techniques in customer relationship management (CRM). It first discusses the importance and goals of CRM, including understanding customer behavior to improve acquisition, retention, loyalty and profitability. It then classifies CRM into operational and analytical approaches. The document proposes a classification framework for AI techniques in CRM, including customer identification, attraction, retention and development. It describes common AI models like association, classification, clustering, forecasting and regression. Finally, it discusses using this framework to systematically review how AI can make CRM more effective.
This published paper aggregates and segments the financial projections of client organizations in multiple industries derived by improving master data quality across the enterprise.
This 3-page document provides an executive summary of a report on how AI is transforming the customer experience. It discusses how AI will become ubiquitous in the next 5 years and profoundly shape interactions with companies through technologies like chatbots and augmented reality. It also outlines some of the key challenges AI poses for customer experience, such as new interaction models, information asymmetry, and the amplification of biases. The summary concludes by emphasizing the need for business leaders to establish principles to ensure AI is developed and applied in a customer-centric manner.
IRJET- Survey on Facial Feature Detection Methods in Offline Recommendation b...IRJET Journal
This document discusses using facial feature extraction for offline recommendation systems. It proposes using age and gender detection from facial images as input for recommendation algorithms without accessing private user data. Age and gender can help categorize users into basic demographic groups associated with different interests. The paper outlines assumptions for the system, including having a camera and processing power for image analysis. It describes common methods for facial feature extraction and age/gender classification, including using global, local, and hybrid facial features. One proposed approach uses multi-level local phase quantization and support vector machines to classify gender from images. The findings could enable targeted recommendations in physical spaces like stores based on visual analysis of users.
Value proposition of analytics in P&C insuranceGregg Barrett
The document provides an overview of the value of analytics in the property and casualty (P&C) insurance industry. It discusses the challenges facing the industry and how analytics can help insurers address these challenges. The document is divided into six sections that cover topics such as the impact of analytics across the insurance life cycle, the value of data and analytics, and considerations for implementing analytics and managing big data. Organizations that effectively adopt analytics are shown to achieve greater growth and returns. While analytics provides opportunities, the document also notes challenges such as developing a data-driven culture and addressing privacy and ethical issues that can arise from certain data collection and analytic practices.
Fairness-aware Machine Learning: Practical Challenges and Lessons Learned (KD...Krishnaram Kenthapadi
Researchers and practitioners from different disciplines have highlighted the ethical and legal challenges posed by the use of machine learned models and data-driven systems, and the potential for such systems to discriminate against certain population groups, due to biases in algorithmic decision-making systems. This tutorial presents an overview of algorithmic bias / discrimination issues observed over the last few years and the lessons learned, key regulations and laws, and evolution of techniques for achieving fairness in machine learning systems. We motivate the need for adopting a "fairness by design" approach (as opposed to viewing algorithmic bias / fairness considerations as an afterthought), when developing machine learning based models and systems for different consumer and enterprise applications. Then, we focus on the application of fairness-aware machine learning techniques in practice by presenting non-proprietary case studies from different technology companies. Finally, based on our experiences working on fairness in machine learning at companies such as Facebook, Google, LinkedIn, and Microsoft, we present open problems and research directions for the data mining / machine learning community.
Artificial intelligence, customer journeys, and paid analytics
Quest to be more data-centric and insights-driven
Data-driven CMOs drive omnichannel customer intelligence
Companies turn to paid analytics for enhanced capabilities
The power of now: customer journey analytics rely on integrated data
Harnessing AI for more insight-driven marketing and better customer experiences
EXTERNAL AND INTERNAL ENVIRONMENTAL ANALYSIS2EXTERNAL AND INTER.docxgitagrimston
EXTERNAL AND INTERNAL ENVIRONMENTAL ANALYSIS 2
EXTERNAL AND INTERNAL ENVIRONMENTAL ANALYSIS 12
External and Internal Environmental Analysis
UNIVERSITY OF PHOENIX
STR/581
Running head: EXTERNAL AND INTERNAL ENVIRONMENTAL ANALYSIS 1
External and Internal Environmental Analysis
Weight Loss Singles Dating Agency is an online dating service tailored to provide companionship services to individuals over the age of 18 who have undergone any form of weight loss surgery. For effective strategic formulation of implementing this new business, it is critical that the company’s external and internal environmental environments are analyzed to ensure competitive advantage is achieved. The following is the analysis of the external and internal environments for the company, its competitive position and possibilities as well as the organizational structure and impact on the organizational performance.
Environmental Analysis
To perform an efficient external and internal environmental analysis it is imperative that the researcher scrutinize all factors affecting implementation of the proposed business. These factors include:
· Evaluate strategic position of Weight Loss Singles Dating Agency
· Assess the online dating market
· Assess proposed business of online dating industry
· Assess the competition
External Environmental Analysis
To provide conclusions needed for successful implementation the researcher will commence with external environmental investigation. Major factors to explore include:
· Barriers to entry
· Threats of new entrants
· Threats of substitutes
· Rivalry between existing firms and
· The power of buyers and suppliers
Barriers to Entry
The company will have very few competitors in contention. Currently, the researcher has discovered that a strong brand recognition does not currently exist as this is a relatively new niche market. Start-up costs will be kept at a minimum as investors will also provide employee assistance in operations. However, increased proposed regulations on the online dating industry will prove to be a challenge when implementing the business model. Currently the industry is not regulated by Federal law, but many US states have proposed bills in legislature to increase protection of users on online dating sites through mandating background checks of all customers (Internet Business Law Services, 2015).
To ensure that barriers to entry will be kept at a minimum the organization will foster a third party association with an outside research firm. This partnership will aid in the determination of illegitimate online dating companies as a strategy for ensuring privacy and integrity concerns (Marsden, 2010). The company will also apply for membership with organizations such as Online Dating Association and Global Dating Insights for the purpose of being recognized as a legitimate business that takes responsibility to maintain high standards in the industry (The Online Dating Association, 2015). This ...
Generative AI has the potential to transform recruitment by automating processes like content creation and screening, improving decision making with data insights, and reducing biases. Its applications could address recruitment challenges related to efficiency, biases, and scalability. However, ethical considerations around data privacy, algorithmic bias, and human-AI collaboration are important to ensure accountability and fairness. With careful integration and oversight, Generative AI may help modernize recruitment in a way that benefits employers, candidates, and society.
Predicting Corporate Failure - An Application of Discriminate Analysisscmsnoida5
Corporate failure is a serious problem being
confronted by the corporate world. This issue
has been a subject of intensive research and
discussion by economists, bankers, creditors,
equity shareholders, accountants, marketing
and management experts. The present study
aims at developing a model for prediction
of corporate failure on the basis of financial
ratios. The study is based on the data of
selected firms from chemical industry (with
equal number of failed and non failed firms).
The discriminant analysis has been used to
discriminate between failed and non failed
firms. It is concluded that some of the
financial ratios can significantly differentiate
between failed and non failed firms. The
finding will be useful for the banks and other
financial institutions in designing a suitable
credit appraisal and monitoring system for their
loans. This model could guide the policy makers
to prepare an early warning system to avoid
bankruptcy.
This document discusses challenges that organizations face in measuring the revenue impact of social media. It notes that while social media is now ubiquitous, most organizations lack a holistic strategy for measuring its effects. It identifies several barriers that organizations encounter, such as the proliferation of social media accounts, inability to tie social activities to business outcomes, lack of analytics expertise, unreliable data, and inconsistent approaches. The document introduces six approaches organizations can use to measure social media's revenue impact and provides case studies of companies that have implemented measurement strategies. It argues that as social media continues to evolve, new measurement approaches will be needed to provide businesses with better insights.
Altimeter, explores the most effective "ingredients" for measuring the revenue impact of social media, providing insights and case studies from an array of top companies.
Similar to Investigating the Impact of Publicly Naming Biased Performance Results of Commercial AI Products (20)
5th LF Energy Power Grid Model Meet-up SlidesDanBrown980551
5th Power Grid Model Meet-up
It is with great pleasure that we extend to you an invitation to the 5th Power Grid Model Meet-up, scheduled for 6th June 2024. This event will adopt a hybrid format, allowing participants to join us either through an online Mircosoft Teams session or in person at TU/e located at Den Dolech 2, Eindhoven, Netherlands. The meet-up will be hosted by Eindhoven University of Technology (TU/e), a research university specializing in engineering science & technology.
Power Grid Model
The global energy transition is placing new and unprecedented demands on Distribution System Operators (DSOs). Alongside upgrades to grid capacity, processes such as digitization, capacity optimization, and congestion management are becoming vital for delivering reliable services.
Power Grid Model is an open source project from Linux Foundation Energy and provides a calculation engine that is increasingly essential for DSOs. It offers a standards-based foundation enabling real-time power systems analysis, simulations of electrical power grids, and sophisticated what-if analysis. In addition, it enables in-depth studies and analysis of the electrical power grid’s behavior and performance. This comprehensive model incorporates essential factors such as power generation capacity, electrical losses, voltage levels, power flows, and system stability.
Power Grid Model is currently being applied in a wide variety of use cases, including grid planning, expansion, reliability, and congestion studies. It can also help in analyzing the impact of renewable energy integration, assessing the effects of disturbances or faults, and developing strategies for grid control and optimization.
What to expect
For the upcoming meetup we are organizing, we have an exciting lineup of activities planned:
-Insightful presentations covering two practical applications of the Power Grid Model.
-An update on the latest advancements in Power Grid -Model technology during the first and second quarters of 2024.
-An interactive brainstorming session to discuss and propose new feature requests.
-An opportunity to connect with fellow Power Grid Model enthusiasts and users.
Ocean lotus Threat actors project by John Sitima 2024 (1).pptxSitimaJohn
Ocean Lotus cyber threat actors represent a sophisticated, persistent, and politically motivated group that poses a significant risk to organizations and individuals in the Southeast Asian region. Their continuous evolution and adaptability underscore the need for robust cybersecurity measures and international cooperation to identify and mitigate the threats posed by such advanced persistent threat groups.
Taking AI to the Next Level in Manufacturing.pdfssuserfac0301
Read Taking AI to the Next Level in Manufacturing to gain insights on AI adoption in the manufacturing industry, such as:
1. How quickly AI is being implemented in manufacturing.
2. Which barriers stand in the way of AI adoption.
3. How data quality and governance form the backbone of AI.
4. Organizational processes and structures that may inhibit effective AI adoption.
6. Ideas and approaches to help build your organization's AI strategy.
Threats to mobile devices are more prevalent and increasing in scope and complexity. Users of mobile devices desire to take full advantage of the features
available on those devices, but many of the features provide convenience and capability but sacrifice security. This best practices guide outlines steps the users can take to better protect personal devices and information.
Fueling AI with Great Data with Airbyte WebinarZilliz
This talk will focus on how to collect data from a variety of sources, leveraging this data for RAG and other GenAI use cases, and finally charting your course to productionalization.
Generating privacy-protected synthetic data using Secludy and MilvusZilliz
During this demo, the founders of Secludy will demonstrate how their system utilizes Milvus to store and manipulate embeddings for generating privacy-protected synthetic data. Their approach not only maintains the confidentiality of the original data but also enhances the utility and scalability of LLMs under privacy constraints. Attendees, including machine learning engineers, data scientists, and data managers, will witness first-hand how Secludy's integration with Milvus empowers organizations to harness the power of LLMs securely and efficiently.
Digital Marketing Trends in 2024 | Guide for Staying AheadWask
https://www.wask.co/ebooks/digital-marketing-trends-in-2024
Feeling lost in the digital marketing whirlwind of 2024? Technology is changing, consumer habits are evolving, and staying ahead of the curve feels like a never-ending pursuit. This e-book is your compass. Dive into actionable insights to handle the complexities of modern marketing. From hyper-personalization to the power of user-generated content, learn how to build long-term relationships with your audience and unlock the secrets to success in the ever-shifting digital landscape.
How to Interpret Trends in the Kalyan Rajdhani Mix Chart.pdfChart Kalyan
A Mix Chart displays historical data of numbers in a graphical or tabular form. The Kalyan Rajdhani Mix Chart specifically shows the results of a sequence of numbers over different periods.
For the full video of this presentation, please visit: https://www.edge-ai-vision.com/2024/06/building-and-scaling-ai-applications-with-the-nx-ai-manager-a-presentation-from-network-optix/
Robin van Emden, Senior Director of Data Science at Network Optix, presents the “Building and Scaling AI Applications with the Nx AI Manager,” tutorial at the May 2024 Embedded Vision Summit.
In this presentation, van Emden covers the basics of scaling edge AI solutions using the Nx tool kit. He emphasizes the process of developing AI models and deploying them globally. He also showcases the conversion of AI models and the creation of effective edge AI pipelines, with a focus on pre-processing, model conversion, selecting the appropriate inference engine for the target hardware and post-processing.
van Emden shows how Nx can simplify the developer’s life and facilitate a rapid transition from concept to production-ready applications.He provides valuable insights into developing scalable and efficient edge AI solutions, with a strong focus on practical implementation.
Salesforce Integration for Bonterra Impact Management (fka Social Solutions A...Jeffrey Haguewood
Sidekick Solutions uses Bonterra Impact Management (fka Social Solutions Apricot) and automation solutions to integrate data for business workflows.
We believe integration and automation are essential to user experience and the promise of efficient work through technology. Automation is the critical ingredient to realizing that full vision. We develop integration products and services for Bonterra Case Management software to support the deployment of automations for a variety of use cases.
This video focuses on integration of Salesforce with Bonterra Impact Management.
Interested in deploying an integration with Salesforce for Bonterra Impact Management? Contact us at sales@sidekicksolutionsllc.com to discuss next steps.
GraphRAG for Life Science to increase LLM accuracyTomaz Bratanic
GraphRAG for life science domain, where you retriever information from biomedical knowledge graphs using LLMs to increase the accuracy and performance of generated answers
How to Get CNIC Information System with Paksim Ga.pptxdanishmna97
Pakdata Cf is a groundbreaking system designed to streamline and facilitate access to CNIC information. This innovative platform leverages advanced technology to provide users with efficient and secure access to their CNIC details.
OpenID AuthZEN Interop Read Out - AuthorizationDavid Brossard
During Identiverse 2024 and EIC 2024, members of the OpenID AuthZEN WG got together and demoed their authorization endpoints conforming to the AuthZEN API
Ivanti’s Patch Tuesday breakdown goes beyond patching your applications and brings you the intelligence and guidance needed to prioritize where to focus your attention first. Catch early analysis on our Ivanti blog, then join industry expert Chris Goettl for the Patch Tuesday Webinar Event. There we’ll do a deep dive into each of the bulletins and give guidance on the risks associated with the newly-identified vulnerabilities.
HCL Notes and Domino License Cost Reduction in the World of DLAUpanagenda
Webinar Recording: https://www.panagenda.com/webinars/hcl-notes-and-domino-license-cost-reduction-in-the-world-of-dlau/
The introduction of DLAU and the CCB & CCX licensing model caused quite a stir in the HCL community. As a Notes and Domino customer, you may have faced challenges with unexpected user counts and license costs. You probably have questions on how this new licensing approach works and how to benefit from it. Most importantly, you likely have budget constraints and want to save money where possible. Don’t worry, we can help with all of this!
We’ll show you how to fix common misconfigurations that cause higher-than-expected user counts, and how to identify accounts which you can deactivate to save money. There are also frequent patterns that can cause unnecessary cost, like using a person document instead of a mail-in for shared mailboxes. We’ll provide examples and solutions for those as well. And naturally we’ll explain the new licensing model.
Join HCL Ambassador Marc Thomas in this webinar with a special guest appearance from Franz Walder. It will give you the tools and know-how to stay on top of what is going on with Domino licensing. You will be able lower your cost through an optimized configuration and keep it low going forward.
These topics will be covered
- Reducing license cost by finding and fixing misconfigurations and superfluous accounts
- How do CCB and CCX licenses really work?
- Understanding the DLAU tool and how to best utilize it
- Tips for common problem areas, like team mailboxes, functional/test users, etc
- Practical examples and best practices to implement right away
UiPath Test Automation using UiPath Test Suite series, part 6DianaGray10
Welcome to UiPath Test Automation using UiPath Test Suite series part 6. In this session, we will cover Test Automation with generative AI and Open AI.
UiPath Test Automation with generative AI and Open AI webinar offers an in-depth exploration of leveraging cutting-edge technologies for test automation within the UiPath platform. Attendees will delve into the integration of generative AI, a test automation solution, with Open AI advanced natural language processing capabilities.
Throughout the session, participants will discover how this synergy empowers testers to automate repetitive tasks, enhance testing accuracy, and expedite the software testing life cycle. Topics covered include the seamless integration process, practical use cases, and the benefits of harnessing AI-driven automation for UiPath testing initiatives. By attending this webinar, testers, and automation professionals can gain valuable insights into harnessing the power of AI to optimize their test automation workflows within the UiPath ecosystem, ultimately driving efficiency and quality in software development processes.
What will you get from this session?
1. Insights into integrating generative AI.
2. Understanding how this integration enhances test automation within the UiPath platform
3. Practical demonstrations
4. Exploration of real-world use cases illustrating the benefits of AI-driven test automation for UiPath
Topics covered:
What is generative AI
Test Automation with generative AI and Open AI.
UiPath integration with generative AI
Speaker:
Deepak Rai, Automation Practice Lead, Boundaryless Group and UiPath MVP
AI 101: An Introduction to the Basics and Impact of Artificial IntelligenceIndexBug
Imagine a world where machines not only perform tasks but also learn, adapt, and make decisions. This is the promise of Artificial Intelligence (AI), a technology that's not just enhancing our lives but revolutionizing entire industries.
WeTestAthens: Postman's AI & Automation Techniques
Investigating the Impact of Publicly Naming Biased Performance Results of Commercial AI Products
1. Actionable Auditing:
Investigating the Impact of Publicly Naming Biased Performance Results of
Commercial AI Products
Inioluwa Deborah Raji
University of Toronto
27 King’s College Cir
Toronto, Ontario, Canada, M5S 3H7
deborah.raji@mail.utoronto.com
Joy Buolamwini
Massachusetts Institute of Technology
77 Massachusetts Ave
Cambridge, Massachusetts, 02139
joyab@mit.edu
Abstract
Although algorithmic auditing has emerged as a key
strategy to expose systematic biases embedded in soft-
ware platforms, we struggle to understand the real-
world impact of these audits, as scholarship on the
impact of algorithmic audits on increasing algorith-
mic fairness and transparency in commercial systems
is nascent. To analyze the impact of publicly naming
and disclosing performance results of biased AI sys-
tems, we investigate the commercial impact of Gender
Shades, the first algorithmic audit of gender and skin
type performance disparities in commercial facial anal-
ysis models. This paper 1) outlines the audit design
and structured disclosure procedure used in the Gen-
der Shades study, 2) presents new performance metrics
from targeted companies IBM, Microsoft and Megvii
(Face++) on the Pilot Parliaments Benchmark (PPB)
as of August 2018, 3) provides performance results on
PPB by non-target companies Amazon and Kairos and,
4) explores differences in company responses as shared
through corporate communications that contextualize
differences in performance on PPB. Within 7 months
of the original audit, we find that all three targets re-
leased new API versions. All targets reduced accuracy
disparities between males and females and darker and
lighter-skinned subgroups, with the most significant up-
date occurring for the darker-skinned female subgroup,
that underwent a 17.7% - 30.4% reduction in error be-
tween audit periods. Minimizing these disparities led to
a 5.72% to 8.3% reduction in overall error on the Pi-
lot Parliaments Benchmark (PPB) for target corporation
APIs. The overall performance of non-targets Amazon
and Kairos lags significantly behind that of the targets,
with error rates of 8.66% and 6.60% overall, and error
rates of 31.37% and 22.50% for the darker female sub-
group, respectively.
Introduction
An algorithmic audit involves the collection and analysis of
outcomes from a fixed algorithm or defined model within
a system. Through the stimulation of a mock user popula-
tion, these audits can uncover problematic patterns in mod-
els of interest. Targeted public algorithmic audits provide
Copyright c 2019, Association for the Advancement of Artificial
Intelligence (www.aaai.org). All rights reserved.
one mechanism to incentivize corporations to address the al-
gorithmic bias present in data-centric technologies that con-
tinue to play an integral role in daily life, from governing
access to information and economic opportunities to influ-
encing personal freedoms (Julia Angwin and Kirchner 2016;
Jakub Mikians 2012; Aniko Hannak and Wilson 2017;
Edelman and Luca 2014).
However, researchers who engage in algorithmic audits
risk breaching company Terms of Service, the Computer
Fraud and Abuse Act (CFAA) or ACM ethical practices as
well as face uncertainty around hostile corporate reactions.
Given these risks, much algorithmic audit work has focused
on goals to gauge user awareness of algorithmic bias (Es-
lami et al. 2017; Kevin Hamilton and Sandvig 2015) or eval-
uate the impact of bias on user behaviour and outcomes
(Gary Soeller and Wilson 2016; Juhi Kulshrestha 2017;
Edelman and Luca 2014), instead of directly challenging
companies to change commercial systems. Research on the
real-world impact of an algorithmic audit is thus needed to
inform strategies on how to engage corporations produc-
tively in addressing algorithmic bias. The Buolamwini &
Gebru Gender Shades study (Buolamwini and Gebru 2018),
which investigated the accuracy of commercial gender clas-
sification services, provides an apt case study to explore au-
dit design and disclosure practices that engage companies
in making concrete process and model improvements to ad-
dress classification bias in their offerings.
Related Work
Corporations and Algorithmic Accountability
As more artificial intelligence (AI) services become main-
stream and harmful societal impacts become increasingly
apparent (Julia Angwin and Kirchner 2016; Jakub Mikians
2012), there is a growing need to hold AI providers ac-
countable. However, outside of the capitalist motivations of
economic benefit, employee satisfaction, competitive advan-
tage, social pressure, and recent legal developments like the
EU General Data Protection Regulation, corporations still
have little incentive to disclose details about their systems
(Diakopoulos 2016; Burrell 2016; Sandra Wachter and Rus-
sell 2018). Thus external pressure remains a necessary ap-
proach to increase transparency and address harmful model
bias.
2. If we take the framing of algorithmic bias as a software
defect or bug that poses a threat to user dignity or access
to opportunity (Tram`er et al. 2015), then we can anticipate
parallel challenges to that faced in the field of information
security, where practitioners regularly address and commu-
nicate threats to user safety. The National Computer Emer-
gency Readiness Team (CERT) promotes a strict procedure
named ”Coordinated Vulnerability Disclosures (CVD)” to
inform corporations of externally identified cyber security
threats in a way that is non-antagonistic, respectful of gen-
eral public awareness and careful to guard against corporate
inaction (Allen D. Householder and King 2017). CVDs out-
line the urgent steps of discovery, reporting, validation and
triage, remediation and then subsequent public awareness
campaigns and vendor re-deployment of a system identified
internally or externally to pose a serious cyber threat. A sim-
ilar ”Coordinated Bias Disclosure” procedure could support
action-driven corporate disclosure practices to address algo-
rithmic bias as well.
Black Box Algorithmic Audits
For commercial systems, the audit itself is characterized as
a ”black box audit”, where the direct or indirect influence of
input features on classifier accuracy or outcomes is inferred
through the evaluation of a curated benchmark (Philip Adler
and Venkatasubramanian 2018; Riccardo Guidotti and Gi-
annotti 2018). Benchmark test sets like FERET (P.J. Phillips
and Rauss 2000) and the Facial Recognition Vendor Test
(FRVT) from the National Institute of Standards and Tech-
nology (NIST) (Mei Ngan and Grother 2015) are of partic-
ular interest, as examples specific to establishing policy and
legal restrictions around mitigating bias in facial recognition
technologies.
In several implemented audit studies, vendor names are
kept anonymous (Brendan F. Klare and Jain 2012) or the
scope is scaled down to a single named target (Snow 2018;
Le Chen and Wilson 2015; Juhi Kulshrestha 2017). The for-
mer fails to harness public pressure and the latter fails to
capture the competitive dynamics of a multi-target audit -
thus reducing the impetus for corporate reactions to those
studies.
Gender Shades
The Gender Shades study differs from these previous cases
as an external and multi-target black box audit of com-
mercial machine learning Application Program Interfaces
(APIs), scoped to evaluating the facial analysis task of bi-
nary gender classification (Buolamwini and Gebru 2018).
The contribution of the work is two-fold, serving to intro-
duce the gender and skin type balanced Pilot Parliaments
Benchmark (PPB) and also execute an intersectional de-
mographic and phenotypic evaluation of face-based gen-
der classification in commercial APIs. The original authors
consider each API’s model performance given the test im-
age attributes of gender, reduced to the binary categories of
male or female, as well as binary Fitzpatrick score, a nu-
merical classification schema for human skin type evaluated
by a dermatologist, and grouped into classes of lighter and
darker skin types. The audit then evaluates model perfor-
mance across these unitary subgroups (i.e. female or darker)
in addition to intersectional subgroups (i.e darker female),
revealing large disparities in subgroup classification accu-
racy particularly across intersectional groups like darker fe-
male, darker male, lighter female and lighter male.
Analysis of Gender Shades Audit
Gender Shades Coordinated Bias Disclosure
In the Gender Shades study, the audit entity is indepen-
dent of target corporations or its competitors and serves as
a neutral ’third-party’ auditor, similar to the expectation for
corporate accounting auditing committees (Allen D. House-
holder and King 2017).
This neutrality enabled auditors to approach audited cor-
porations systematically, following a procedure sequentially
outlined below that closely mirrors key recommendations
for coordinated vulnerability disclosures (CVDs) in infor-
mation security (Allen D. Householder and King 2017).
1. Documented Vulnerability Discovery - A stated objec-
tive of the Gender Shades study is to document audit out-
comes from May 2017 to expose performance vulnera-
bilities in commercial facial recognition products (Buo-
lamwini and Gebru 2018).
2. Defined Corporate Response Period with Limited
Anonymomized Release to Audit Targets - The Gender
Shades paper (without explicit company references) was
sent to Microsoft, IBM and Face++ on December 19th
2017 (Buolamwini 2017), giving companies prior notice
to react before a communicated public release date, while
maintaining the strict privacy of other involved stakehold-
ers.
3. Unrestricted Public Release Including Named Audit
Targets - On February 9th, 2018, ”Facial Recognition Is
Accurate, if You’re a White Guy” , an article by Steve
Lohr in the technology section of The New York Times
is among the first public mentions of the study (Buo-
lamwini 2017; 2018), and links to the published version
of the study in Proceedings of Machine Learning Re-
search, with explicit company references. This follows
CVD procedures around alerting the public of corporate
vulnerabilities with explicit culprit references, following
a particular grace period in which companies are allowed
to react before wider release. The Gender Shades public
launch, accompanied by a video, summary visualizations
and a website further prompts public, academic and cor-
porate audiences - technical and non-technical alike - to
be exposed to the issue and respond. Finally, the paper
was presented on February 24th 2018 with explicit com-
pany references at the FAT* conference to an audience of
academics, industry stakeholders and policymakers (Buo-
lamwini 2017).
4. Joint Public Release of Communications and Updates
from Corporate Response Period - Even if the issue is
resolved, CVD outlines a process to still advance with the
public release while also reporting corporate communica-
tions and updates from the response period. In the case
3. Figure 1: Gender Shades audit process overview (Buo-
lamwini and Gebru 2018).
of Gender Shades, the co-author presented and linked to
IBMs updated API results at the time of the public release
of the initial study (Buolamwini 2017).
Gender Shades Audit Design
The Gender Shades paper contributed to the computer vision
community the Pilot Parliaments Benchmark (PPB). PPB
is an example of what we call a ”user-representative” test
set, meaning the benchmark does not have proportional de-
mographic distribution of the intended user population but
representative inclusion of the diversity of that group. With
equal representation of each distinct subgroup of the user
population regardless of the percentage at which that popu-
lation is present in the sample of users, we can thus evaluate
for equitable model performance across subgroups. Similar
to the proposed error profiling under the Unwarranted As-
sociations framework (Tram`er et al. 2015), algorithmic un-
fairness is evaluated by comparing classification accuracy
across identified subgroups in the user-representative test set
(see Figure 1).
Another key element of the audit design is that the audit
targets are commercial machine learning Application Pro-
gram Interfaces (APIs). The auditors thus mirror the be-
haviour of a single developer user for a commercial API
platform that supports the creation of applications for the
end user. Therefore, the actor being puppeted (the developer)
has control of the application being used by the end-user
and is at risk of propagating bias unto the end-users of their
subsequent products. This is analogous to the ”sock pup-
pet” algorithmic audit model (Christian Sandvig and Lang-
bort 2014) in that we pose as a puppet user of the API plat-
form and interact with the API in the way a developer would.
However, as this is a puppet that influences the end user ex-
perience, we label them ”carrier puppets”, acknowledging
that, rather than evaluating a final state, we are auditing the
bias detected in an intermediary step that can carry bias for-
ward towards end users (see Figure 2).
Methodology
The design of this study is closely modeled after that of Gen-
der Shades. Target corporations were selected from the orig-
inal study, which cites considerations such as the platform
Figure 2: ”Carrier puppet” audit framework overview.
market share, the availability of desired API functions and
overall market influence as driving factors in the decision
to select Microsoft, IBM and Face++ (Buolamwini and Ge-
bru 2018). Non-target corporation Kairos was selected be-
cause of the company’s public engagement with the Gender
Shades study specifically and the topic of intersectional ac-
curacy in general after the audit release (Brackeen 2018a;
2018b). Non-target corporation Amazon was selected fol-
lowing the revelation of the active use and promotion of its
facial recognition technology in law enforcement (Cagle and
Ozer 2018).
The main factor in analysis, the follow up audit, closely
follows the procedure for the initial Gender Shades study.
We calculated the subgroup classification error, as defined
below, to evaluate disparities in model performance across
identified subgroups, enabling direct comparison between
follow-up results and initial audit results.
Subgroup Classification Error. Given data set D =
(X, Y, C), a given sample input di from D belongs to a sub-
group S, which is a subset of D defined by the protected
attributes X. We define black box classifier g : X, Y → c,
which returns a prediction c from the attributes xi and yi of
a given sample input di from D. If a prediction is not pro-
duced (i.e. face not detected), we omit the result from our
calculations.
We thus define err(S) be the error of the classifier g for
members di of subgroup S to be as follows:
1 − P(g(xi, yi) = Ci|di ∈ S)
To contextualize audit results and examine language
themes used post-audit, we considered written communica-
tions for all mentioned corporations. This includes exclu-
sively corporate blog posts and official press releases, with
the exception of media published corporate statements, such
as an op-ed by the Kairos CEO published in TechCrunch
(Brackeen 2018b). Any past and present website copy or
Software Developer Kit documentation was also considered
when determining alignment with identified themes, though
this did not factor greatly into the results.
Performance Results
With the results of the follow up audit and original Gen-
der Shades outcomes, we first analyze the differences be-
tween the performance of the targeted platforms in the orig-
inal study and compare it to current target API performance.
Next, we look at non-target corporations Kairos and Ama-
zon, which were not included in the Gender Shades study
and compare their current performance to that of targeted
platforms.
4. Table 1: Overall Error on Pilot Parliaments Benchmark, August 2018 (%)
Company All Females Males Darker Lighter DF DM LF LM
Target Corporations
Face ++ 1.6 2.5 0.9 2.6 0.7 4.1 1.3 1.0 0.5
MSFT 0.48 0.90 0.15 0.89 0.15 1.52 0.33 0.34 0.00
IBM 4.41 9.36 0.43 8.16 1.17 16.97 0.63 2.37 0.26
Non-Target Corporations
Amazon 8.66 18.73 0.57 15.11 3.08 31.37 1.26 7.12 0.00
Kairos 6.60 14.10 0.60 11.10 2.80 22.50 1.30 6.40 0.00
Table 2: Overall Error Difference Between August 2018 and May 2017 PPB Audit (%)
Company All Females Males Darker Lighter DF DM LF LM
Face ++ -8.3 -18.7 0.2 -13.9 -3.9 -30.4 0.6 -8.5 -0.3
MSFT -5.72 -9.70 -2.45 -12.01 -0.45 -19.28 -5.67 -1.06 0.00
IBM -7.69 -10.74 -5.17 -14.24 -1.93 -17.73 -11.37 -4.43 -0.04
The reported follow up audit was done on August 21,
2018, for all corporations in both cases. Summary Table 1
and Table 2 show percent error on misclassified faces of
all processed faces, with undetected faces being discounted.
Calculation details are outlined in the definition for Sub-
group Classification Error and error differences are calcu-
lated by taking August 2018 error (%) and subtracting May
2017 error (%). DF is defined as darker female subgroup,
DM is darker male, LM is lighter male and LF is lighter fe-
male.
Target Corporation Key Findings
The target corporations from the Gender Shades study all re-
leased new API versions, with a reduction in overall error on
the Pilot Parliamentary Benchmark by 5.7%, 8.3% and 7.7%
respectively for Microsoft, Face++ and IBM. Face++ took
the most days to release their new API in 190 days (Face++
2018), while IBM was the first to release a new API version
in 66 days (Puri 2018), with Microsoft updating their prod-
uct the day before Face++, in 189 days (Roach 2018). All
targeted classifiers in post-audit releases have their largest
error rate for the darker females subgroup and the lowest
error rate for the lighter males subgroup. This is consistent
with 2017 audit trends, barring Face++ which had the lowest
error rate for darker males in May 2017.
The following is a summary of substantial performance
changes across demographic and phenotypic classes, as well
as their intersections, after API updates :
• Greater reduction in error for female faces ( 9.7% - 18.7%
reduction in subgroup error) than male faces ( 0.2% -
5.17% reduction in error) .
• Greater reduction in error for darker faces ( 12.01% -
14.24% reduction in error) than for lighter faces ( 0.45%
- 3.9% reduction in error).
• Lighter males are the least improved subgroup ( 0% -
0.3% reduction in error)
• Darker females are the most improved subgroup (17.7% -
30.4% reduction in error)
• If we define the error gap to be the error difference be-
tween worst and best performing subgroups for a given
API product, IBM reduced the error gap from 34.4% to
16.71% from May 2017 to August 2018. In the same pe-
riod, Microsoft closed a 20.8% error gap to a 1.52% error
difference, and Face++ went from a 33.7% error gap to a
3.6% error gap.
Non-Target Corporation Key Findings
Non-target corporations Kairos and Amazon have overall er-
ror rates of 6.60% and 8.66% respectively. These are the
worst current performances of the companies analyzed in
the follow up audit. Nonetheless, when comparing to the
previous May 2017 performance of target corporations, the
Kairos and Amazon error rates are lower than the former
error rates of IBM (12.1%) and Face++ (9.9%), and only
slightly higher than Microsofts performance (6.2%) from the
initial study. Below is a summary of key findings for non-
target corporations:
• Kairos and Amazon perform better on male faces than fe-
male faces, a trend also observed in (Buolamwini and Ge-
bru 2018; Mei Ngan and Grother 2015).
• Kairos and Amazon perform better on lighter faces than
darker faces, a trend also observed in (Buolamwini and
Gebru 2018; Jonathon Phillips and OToole 2011).
• Kairos (22.5% error) and Amazon (31.4% error) have
the current worst performance for the darker female sub-
group.
• Kairos and Amazon (both 0.0% error) have the current
best performance for the lighter male subgroup.
• Kairos has an error gap of 22.5% between highest and
lowest accuracy intersectional subgroups, while Amazon
has an error gap of 31.37%.
5. Discussion
Given a clear understanding of the Gender Shades study pro-
cedure and follow up audit metrics, we are able to reflect on
corporate reactions in the context of these results, and eval-
uate the progress made by this audit in influencing corporate
action to address concerns around classification bias.
Reduced Performance Disparities Between
Intersectional User Subgroups
Building on Crenshaw’s 1989 research on the limitations
of only considering single axis protected groups in anti-
discrimination legislation (Crenshaw 1989), a major focus
of the Gender Shades study is championing the relevance of
intersectional analysis in the domain of human-centered AI
systems. IBM and Microsoft, who both explicitly reference
Gender Shades in product update releases, claim intersec-
tional model improvements on their gender classifier (Puri
2018; Roach 2018). These claims are substantiated by the
results of the August 2018 follow up audit, which reveals
universal improvement across intersectional subgroups for
all targeted corporations. We also see that the updated re-
leases of target corporations mostly impact the least accurate
subgroup (in this case, darker females). Although post-audit
performance for this subgroup is still the worst relative to
other intersectional subgroups across all platforms, the gap
between this subgroup and the best performing subgroup -
consistently lighter males - reduces significantly after cor-
porate API update releases.
Additionally, with a 5.72% to 8.3% reduction in overall
error on the Pilot Parliaments Benchmark (PPB) for target
corporations, we demonstrate that minimizing subgroup per-
formance disparities does not jeopardize overall model per-
formance but rather improves it, highlighting the alignment
of fairness objectives to the commercial incentive of im-
proved qualitative and quantitative accuracy. This key result
highlights an important critique of the current model evalua-
tion practice of using a subset of the model training data for
testing, by demonstrating the functional value in testing the
model on a separately defined ”user representative” test set.
Corporate Prioritization
Although the original study (Buolamwini and Gebru 2018)
expresses the concern that potential physical limitations of
the image quality and illumination of darker skinned sub-
jects may be contributing to the higher error rate for that
group, we can see through the 2018 performance results that
these challenges can be overcome. Within 7 months, all tar-
geted corporations were able to significantly reduce error
gaps in the intersectional performance of their commercial
APIs, revealing that if prioritized, the disparities in perfor-
mance between intersectional subgroups can be addressed
and minimized in a reasonable amount of time.
Several factors may have contributed to this increased pri-
oritization. The unbiased involvement of multiple compa-
nies may have served to put capitalist pressure on each cor-
poration to address model limitations as not to be left be-
hind or called out. Similarly, increased corporate and con-
sumer awareness on the issue of algorithmic discrimination
and classification bias in particular may have incited ur-
gency in pursuing a product update. This builds on litera-
ture promoting fairness through user awareness and educa-
tion (Kevin Hamilton and Eslami 2014) - aware corporations
can also drastically alter the processes needed to reduce bias
in algorithmic systems.
Emphasis on Data-driven Solutions
These particular API updates appear to be data-driven. IBM
publishes the statement ”AI systems are only as effective
as the data they’re trained on” and both Microsoft and
Kairos publish similar statements (Puri 2018; Roach 2018;
Brackeen 2018a) , implying heavily the claim that data col-
lection and diversification efforts play an important role
in improving model performance across intersectional sub-
groups. This aligns with existing research (Irene Chen and
Sontag 2018) advocating for increasing the diversity of data
as a primary approach to improve fairness outcomes without
compromising on overall accuracy. Nevertheless, the influ-
ence of algorithmic changes, training methodology or spe-
cific details about the exact composition of new training
datasets remain unclear in this commercial context - thus
underscoring the importance of work on open source mod-
els and datasets that can be more thoroughly investigated.
Non-technical Advancements
In addition to technical updates, we observe organizational
and systemic changes within target corporations following
the Gender Shades study. IBM published its ”Principles for
Trust and Transparency” on May 30th 2018 (IBM 2018),
while Microsoft created an ”AI and Ethics in Engineering
and Research (AETHER) Committee, investing in strategies
and tools for detecting and addressing bias in AI systems” on
March 29th, 2018 (Smith 2018). Both companies also cite
their involvement in Partnership for AI, an AI technology
industry consortium, as a means of future ongoing support
and corporate accountability (Puri 2018; Smith 2018).
Implicitly identifying the role of the API as a ”carrier”
of bias to end users, all companies also mention the impor-
tance of developer user accountability, with Microsoft and
IBM speaking specifically to user engagement strategies and
educational material on fairness considerations for their de-
veloper or enterprise clients (Puri 2018; Roach 2018).
Only Microsoft strongly mentions the solution of Diver-
sity & Inclusion considerations in hiring as an avenue to ad-
dress issues[38]. The founder of Kairos specifically claims
his minority identity as personal motivation for participa-
tion in this issue, stating ”I have a personal connection to
the technology,...This resonates with me very personally as
a minority founder in the face recognition space” (Brackeen
2018a; 2018b). A cultural shift in the facial recognition in-
dustry could thus attract and retain those paying increased
attention to the issue due to personal resonance.
Differences between Target and Non Target
Companies
Although prior performance for non-target companies is un-
known, and no conclusions can be made about the rate of
6. product improvements, Kairos and Amazon both perform
more closely to the target corporations’ pre-audit perfor-
mance than their post-audit performance.
Amazon, a large company with an employee count and
revenue comparable to the target corporations IBM and Mi-
crosoft, seems optimistic about the use of facial recognition
technology despite current limitations. In a response to a
targeted ACLU audit of their facial recognition API (Snow
2018), they state explicitly, ”Our quality of life would be
much worse today if we outlawed new technology because
some people could choose to abuse the technology”. On the
other hand, Kairos, a small privately held company not ex-
plicitly referenced in the Gender Shades paper and subse-
quent press discussions, released a public response to the
initial Gender Shades study and seemed engaged in taking
the threat of algorithmic bias quite seriously (Buolamwini
and Gebru 2018).
Despite the varying corporate stances and levels of public
engagement,the targeted audit in Gender Shades was much
more effective in reducing disparities in target products than
non-targeted systems.
Regulatory Communications
We additionally encounter scenarios where civil society or-
ganizations and government entities not explicitly refer-
enced in the Gender Shades paper and subsequent press
discussions publicly reference the results of the audit in
letters, publications and calls to action. For instance, the
Gender Shades study is cited in an ACLU letter to Ama-
zon from shareholders requesting its retreat from selling
and advertising facial recognition technology for law en-
forcement clients (Arjuna Capital 2018). Similar calls for
action to Axon AI by several civil rights groups, as well
as letters from Senator Kamala D. Harris to the EEOC,
FBI and FTC regarding the use of facial recognition in
law enforcement also directly reference the work (Cold-
ewey 2017). Kairos, IBM and Microsoft all agree facial
analysis technology should be restricted in certain con-
texts and demonstrate support for government regulation
of facial recognition technology (IBM 2018; Smith 2018;
Brackeen 2018b). In fact, Microsoft goes so far as to explic-
itly support public regulation (Smith 2018). Thus in addi-
tion to corporate reactions, future work might explore the
engagement of government entities and other stakeholders
beyond corporate entities in response to public algorithmic
audits.
Design Considerations
Several design considerations also present opportunities for
further investigation. As mentioned in Gender Shades, a
consideration of confidence scores on these models is nec-
essary to get a complete view on defining real-world perfor-
mance (Buolamwini and Gebru 2018). For instance, IBM’s
self-reported performance on a replicated version of the
Gender Shades audit claims a 3.46% overall error rate on
their lowest accuracy group of darker females (Puri 2018) -
this result varies greatly from the 16.97% error rate we ob-
serve in our follow up audit. Upon further inspection, we
see that they only include results above a 99% confidence
threshold whereas Gender Shades takes the binary label with
the higher confidence score to be the predicted gender. These
examples demonstrate the need to consider variations in re-
sults due to prediction confidence thresholding in future au-
dit designs.
Another consideration is that the Gender Shades publi-
cation includes all the required information to replicate the
benchmark and test models on PPB images (Buolamwini
and Gebru 2018). It is possible that well performing models
do not truly perform well on other diverse datasets outside of
PPB and have been overfit to optimize their performance on
this particular benchmark. Future work involves evaluation
of these systems on a separate balanced dataset of similar de-
mographic attributes to PPB or making use of metrics such
as balanced error to account for class imbalances in existing
benchmarks.
Additionally, although Face++ appears to be the least en-
gaged or responsive company, a limitation of the survey to
English blog posts and American mainstream media quotes
(Face++ 2018), definitively excludes Chinese media outlets
that would reveal more about the company’s response to the
audit.
Conclusion
Therefore, we can see from this follow-up study that all
target companies reduced classification bias in commercial
APIs following the Gender Shades audit. By highlighting
the issue of classification performance disparities and am-
plifying public awareness, the study was able to motivate
companies to prioritize the issue and yield significant im-
provements within 7 months. When observed in the context
of non-target corporation performance, however, we see that
significant subgroup performance disparities persist. Never-
theless, corporations outside the scope of the study continue
to speak up about the issue of classification bias (Brack-
een 2018b). Even those less implicated are now facing in-
creased scrutiny by civil groups, governments and the con-
sumers as a result of increased public attention to the issue
(Snow 2018). Future work includes the further development
of audit frameworks to understand and address corporate en-
gagement and awareness, improve the effectiveness of algo-
rithmic audit design strategies and formalize external audit
disclosure practices.
Furthermore, while algorithmic fairness may be ap-
proximated through reductions in subgroup error rates or
other performance metrics, algorithmic justice necessitates
a transformation in the development, deployment, oversight,
and regulation of facial analysis technology. Consequently,
the potential for weaponization and abuse of facial analysis
technologies cannot be ignored nor the threats to privacy or
breaches of civil liberties diminished even as accuracy dis-
parities decrease. More extensive explorations of policy, cor-
porate practice and ethical guidelines is thus needed to en-
sure vulnerable and marginalized populations are protected
and not harmed as this technology evolves.
7. References
Allen D. Householder, Garret Wassermann, A. M., and King,
C. 2017. The cert guide to coordinated vulnerability disclo-
sure. Government technical report, Carnegie Mellon Univer-
sity.
Aniko Hannak, Claudia Wagner, D. G. A. M. M. S., and Wil-
son, C. 2017. Bias in online freelance marketplaces: Evi-
dence from taskrabbit and fiverr. In 2017 ACM Conference,
1914–1933. New York, NY, USA: ACM.
Arjuna Capital, As You Sow, C. A. M. J. C. D. S. o. H. D. I.
L. F. . I. S. F. H. P. C. H. I. I. M. S. M. N. C. f. R. I. S. A. M.
T. S. E. G. T. S. G. o. L. W. . C. T. W. M. L. U. S. o. T. U. P.
W. A. M. Z. A. M. 2018. Letter from shareholders to amazon
ceo jeff bezos regarding rekognition.
Brackeen, B. 2018a. Face off: Confronting bias in face recog-
nition ai.
Brackeen, B. 2018b. Facial recognition software is not ready
for use by law enforcement.
Brendan F. Klare, Mark J. Burge, J. C. K. R. W. V. B., and
Jain, A. K. 2012. Face recognition performance: Role of
demographic information. In IEEE Transactions on Infor-
mation Forensics and Security, volume 7, 1789–1801. New
York, NY, USA: IEEE.
Buolamwini, J., and Gebru, T. 2018. Gender shades: Inter-
sectional accuracy disparities in commercial gender classifi-
cation. In Proceedings of Machine Learning Research. Con-
ference on Fairness, Accountability, and Transparency.
Buolamwini, J. 2017. Gender shades.
Buolamwini, J. 2018. When the robot doesnt see dark skin.
Burrell, J. 2016. How the machine thinks: Understanding
opacity in machine learning algorithms. Big Data & Society.
Cagle, M., and Ozer, N. 2018. Amazon teams up with govern-
ment to deploy dangerous new facial recognition technology.
Christian Sandvig, Kevin Hamilton, K. K., and Langbort, C.
2014. Auditing algorithms: Research methods for detecting
discrimination on internet platforms. Data and Discrimina-
tion, Converting Critical Concerns into Productive: A Pre-
conference at the 64th Annual Meeting of the International
Communication Association.
Coldewey, D. 2017. Sen. harris tells federal agencies to get
serious about facial recognition risks.
Crenshaw, K. 1989. Demarginalizing the intersection of race
and sex: A black feminist critique of antidiscrimination doc-
trine, feminist theory and antiracist politics. University of
Chicago Legal Forum 1989(8).
Diakopoulos, N. 2016. Accountability in algorithmic deci-
sion making. Communications of the ACM 59(2):56–62.
Edelman, B., and Luca, M. 2014. Digital discrimination: The
case of airbnb.com. SSRN Electronic Journal.
Eslami, M.; Vaccaro, K.; Karahalios, K.; and Hamilton, K.
2017. Be careful, things can be worse than they appear:
Understanding biased algorithms and users’ behavior around
them in rating platforms. In ICWSM.
Face++. 2018. Notice: newer version of face detect api.
Gary Soeller, Karrie Karahalios, C. S., and Wilson, C. 2016.
Mapwatch: Detecting and monitoring international border
personalization on online maps. J. ACM.
IBM. 2018. Ibm principles for trust and transparency.
Irene Chen, F. D. J., and Sontag, D. 2018. Why is my classi-
fier discriminatory? In arXiv preprint. arXiv.
Jakub Mikians, Lszl Gyarmati, V. E. a. N. L. 2012. Detecting
price and search discrimination on the internet. In Proceed-
ings of the 11th ACM Workshop on Hot Topics in Networks,
HotNets-XI. New York, NY, USA: ACM.
Jonathon Phillips, Fang Jiang, A. N. J. A., and OToole, A. J.
2011. An other-race effect for face recognition algorithms. In
ACM Transactions on Applied Perception (TAP), volume 8.
ACM Press.
Juhi Kulshrestha, Motahhare Eslami, J. M. M. B. Z. S. G.
K. P. G. K. K. 2017. Quantifying search bias: Investigat-
ing sources of bias for political searches in social media. In
Proceedings of the 2017 ACM Conference on Computer Sup-
ported Cooperative Work and Social Computing, 417–432.
New York, NY, USA: ACM.
Julia Angwin, Jeff Larson, S. M., and Kirchner, L. 2016.
Machine bias.
Kevin Hamilton, Karrie Karahalios, C. S., and Eslami, M.
2014. A path to understanding the effects of algorithm aware-
ness. In CHI ’14 Extended Abstracts on Human Factors in
Computing Systems (CHI EA ’14), 631–642. New York, NY,
USA: ACM.
Kevin Hamilton, Motahhare Eslami, A. A. K. K., and Sand-
vig, C. 2015. I always assumed that i wasn’t really that close
to [her]: Reasoning about invisible algorithms in the news
feed. In Proceedings of 33rd Annual ACM Conference on
Human Factors in Computing Systems. New York, NY, USA:
ACM.
Le Chen, A. M., and Wilson, C. 2015. Peeking beneath the
hood of uber. In Proceedings of 2015 ACM Conference, 495–
508. New York, NY, USA: ACM.
Mei Ngan, M. N., and Grother, P. 2015. Face recognition ven-
dor test (frvt) performance of automated gender classification
algorithms. Government technical report, US Department of
Commerce, National Institute of Standards and Technology.
Philip Adler, Casey Falk, S. A. F. T. N. G. R. C. S. B. S., and
Venkatasubramanian, S. 2018. Auditing black-box models
for indirect influence. Knowledge and Information Systems
54(1).
P.J. Phillips, Hyeonjoon Moon, S. R., and Rauss, P. 2000.
The feret evaluation methodology for face-recognition algo-
rithms. In IEEE Transactions on Pattern Analysis and Ma-
chine Intelligence, volume 22, 1090 – 1104. IEEE.
Puri, R. 2018. Mitigating bias in ai models.
Riccardo Guidotti, Anna Monreale, F. T. D. P., and Giannotti,
F. 2018. A survey of methods for explaining black box mod-
els. ACM Computing Surveys 51(5).
Roach, J. 2018. Microsoft improves facial recognition tech-
nology to perform well across all skin tones, genders.
Sandra Wachter, B. M., and Russell, C. 2018. Counterfactual
explanations without opening the black box: Automated de-
cisions and the gdpr. Harvard Journal of Law & Technology
31(2).
Smith, B. 2018. Facial recognition technology: The need for
public regulation and corporate responsibility.
Snow, J. 2018. Amazon’s face recognition falsely matched
28 members of congress with mugshots.
Tram`er, F.; Atlidakis, V.; Geambasu, R.; Hsu, D. J.; Hubaux,
J.-P.; Humbert, M.; Juels, A.; and Lin, H. 2015. Discovering
unwarranted associations in data-driven applications with the
fairtest testing toolkit. CoRR abs/1510.02377.