The document discusses the importance of trustworthy AI in data-driven economies, emphasizing transparency, fairness, and the reduction of information asymmetry between service producers and consumers. It highlights the challenges posed by algorithmic bias, the necessity for robust and responsible AI, and the need for interdisciplinary and international collaboration to address these issues. Key initiatives include the ethical guidelines by the European Commission and the OECD's efforts to map the social impacts of AI applications.

1. 19/12/18
1
- 1Paris - November 2018
Towards Trustworthy AI
Nozha Boujemaa
MIT Tech Review – Innovation Leaders Summit
Research Director at Inria
Vice-Chair of AI HLEG – European Commission
nozha.boujemaa@inria.fr
- 2
Data & Algorithms
• Transparency and trust of such Algorithmic Systems (data & algorithms)
becoming competitiveness factors for Data-driven economy ;
• Data analytics is changing from description of past to predictive and
prescriptive analytics for decision support ;
• Importance of reducing the information asymmetry between the producer
of the digital service and its consumer, be it citizen or professional – B2C or
B2B => civil rights, competition, sovereignty.
« 2 sides of the same coin »
Data Science, Intelligence & Society

2. 19/12/18
2
- 3
Algorithmic systems in every day life
• Some platforms play a role of "prescriber” by directing user traffic:
• Ranking mechanisms (search engine),
• Recommendation mechanisms and content selection
• Opacity of the use made of the personal data
• What about the consent? Is it always respected?
• Credit scoring, Predictive justice, how fair are they ??
⇒ New discrimination between those who know how algorithms work and
who do not
Data Science, Intelligence & Society
- 4
4
Explanation:
Ribeiro et al. 2016, LIME: Why should I trust you?
Explaining the predictions of any classifier
Data Science, Intelligence & Society
Safe AI: Robustness and Explanation
Robustness:
Goodfellow, Shlens and Szegedy 2015, “Explaining and
Harnessing Adversarial Examples”

3. 19/12/18
3
- 5
Algorithmic Systems Bias
Mastering Big Data Technologies: Bias problems could impact data
technologies accuracy and people’s lives
Challenges 1: Data Inputs to an Algorithm
o Poorly selected data, lack of representativeness
o Incomplete, incorrect, or outdated data
o Malicious attack
Challenges 2: The Design of Algorithmic Systems and Machine Learning
o Unintentional perpetuation and promotion of historical biases
o Decision-making systems that assume correlation implies causation
o Poorly designed matching systems
Data Science, Intelligence & Society
- 6
Challenges
• Mastering the accuracy and robustness of Big Data & AI
techniques: bias, diversion/corruption, reproducibility, source
of unintentional discrimination
• Algorithms are encapsulated opinions through decision
parameters and learning data => intention
• Implementing new generation “Transparent-by-Design”:
fairness/equity, loyalty, neutrality => “Value-by-Design”
Data Science, Intelligence & Society

4. 19/12/18
4
- 7
Challenges
§ Trusted AI:
§ Responsible: Compliance with Regulation/Policy and Social Values/Ethics
§ Robust and safe: against bias, corruption, noise, reproducibility etc
§ AI is part of the solution and not only the law! Algorithmic tools to monitor the
behavior of AI technologies
§ Auditability and Responsible-by-Design tools and algorithms for socio-
economic empowerment => Traceability as facilitator
Data Science, Intelligence & Society
- 8
AI International Efforts
High Level Expert Group of the European Commission (AI HLEG)
• 52 independent experts from industry, academia, consumer associations (computer science,
law, ethicist, philosopher, entrepreneur, activist)
o Ethical guidelines for Trustworthy AI (mid-December 218)
o Elaboration of recommendations on AI Policy & Investment (March 2018)
Artificial Intelligence Expert Group at the OECD (AIGO)
• 36 members: OECD governments representatives + Experts : MIT, Harvard, Inria, IEEE, Civil
Society (November 2019)
o map economic / social impacts of AI applications,
o discuss policies that influence adoption of AI and policies to address its consequences
Data Science, Intelligence & Society

5. 19/12/18
5
- 9
Merci de votre attention
Science des données, Intelligence & SociétéScience des données, Intelligence & Société
Need for Interdisciplinary & International efforts
THANK YOU
nozha.boujemaa@inria.fr
@NozhaBoujemaa
Data Science, Artificial Intelligence & Society
- 10