Optimizing Data Synthesis 
and Visualization in Real- 
Time Decision-Making 
Kim Bender, AMS Summer Meeting 
August 2014
“As weather forecaster you will be asked to make decisions 
in situations where you will be under tight time constraints 
and face uncertainty, you will have huge amounts of data 
available, some of it contradictory and most of it not really 
of significance to you, and you will be constantly looking for 
the correct data, which will not normally be at your disposal 
(Gaia & Fontannaz, 2006, p.17)” 
Confidential and Proprietary. Do Not Distribute. 2
Data Synthesis and Visualization in Real-Time 
Decision Making 
• Forecasting and aviation operations are both technological environments 
where data is increasingly being incorporated into operator workstations to 
optimize accuracy and efficiency 
○ Automation 
○ Decision Support Tools 
○ Systems 
Confidential and Proprietary. Do Not Distribute. 3 
• Issues and questions: 
○ How do you ensure the right systems are incorporated? 
○ How do you ensure effective presentation and synthesis of information? 
○ What are the impacts of the system on the human operator, safety, and 
operations? 
• What predisposes these operations to these environments? 
Operators are required to make real-time decisions 
in dynamically changing, uncertain environments
Human Factors: Decision Making in Uncertainty 
• Decisions Making in Uncertainty 
○ Heuristics – intuitive processes based on individual experience and perceptions 
○ Objective data – structured and unstructured data but usually statistical in nature 
• Heuristics mandatory in uncertain environments: Intuition and experience help 
humans make quick effective decisions 
○ Best human forecasters can consistently outperform objective methods- 
Confidential and Proprietary. Do Not Distribute. 4 
Heuristics! 
○ Heuristics influenced by past experience and subjective considerations which can 
be positive or negative 
■ Positive – recognize similar situation or pattern and apply lessons learned when data is 
uncertain 
■ Negative – can introduce bias due to previous negative outcome or experience or it 
can lead to missed event 
□ Previous false alarm, forecasters tends to under forecast next event leading to a missed 
alarm 
□ Previous missed alarm, forecasters tend to over forecast after missed event
Human Factors: Decision Making in Uncertainty 
• Inordinate amounts of objective data for real-time decision-making can be 
problematic when not introduced effectively 
• Operator must mine/culls/filter data to quickly synthesize 
○ Multiple data sources can introduce more uncertainty 
■ Contradictory output consistencies and forecasts 
■ Individual data interpolation and integration/translation differences 
○ Data overload – huge amounts of data 
■ Contradictory data sources from disparate sources and systems 
■ Most data not really of significance for task at hand 
Confidential and Proprietary. Do Not Distribute. 5 
○ Data Visualization 
■ Scanning for correct/relevant data can be time consuming and problematic 
■ Relevant data may not always be available or clear 
■ Multiple systems introduce display and output inconsistencies 
If data synthesis and visualization can be structured 
carefully, could mitigate many significant issues
Evaluating Effective Data Presentation and 
Visualization 
• FAA tests proposed advanced aviation systems, procedures, and concepts 
and/or performs experimental research to test proposed change 
○ Subject Matter Expert (SME) Panels 
Confidential and Proprietary. Do Not Distribute. 6 
○ Simulation Research 
○ Field Research 
• Introduction of new data, information, and/or display in aviation environment 
requires: 
○ Operational Testing and Evaluation using SME’s 
○ Measurement/evaluation of impacts of proposed technologies, displays, or 
concepts on operations, safety, or human operator 
○ SME evaluation of information display
Key Considerations for Data Synthesis and 
Visualization 
• Data introduction and visualization considerations 
○ Data selection, design, display and is key to performance 
○ Only present data that is meaningful and accurate - “too much” data will 
eventually decrease performance by adding ineffective workload 
• Implement a systematic approach to data presentation and visualization 
○ SME and user evaluations during development and operational testing 
○ Disciplined and systematic approach data presentation and visualization display 
characteristics / system integration 
■ Keep systems visually consistent – colors, presentation, warnings, etc. 
■ System integration –data consistency 
Confidential and Proprietary. Do Not Distribute. 7
Key Considerations for Data Synthesis and 
Visualization 
• Experimental research to understand the effects of the system on human 
performance and evaluate alternative approaches 
• Understand human decision making and how operator 
visualizes/interacts with large datasets in operational settings 
○ Understand each operation – all different 
○ Cognitive walkthroughs and walk-through like methods 
• What data should be presented and how much to automate? 
○ Some processes/tasks should not be automated 
○ Keep human-in-the-loop to ensure they are active, involved, and aware 
○ Research and study need to determine best automation candidates 
○ Operational design needs to be structured to tasks 
Confidential and Proprietary. Do Not Distribute. 8
Key Considerations for Data Synthesis and 
Visualization 
• Workstation should make best use of human strengths: 
○ Recognizing patterns 
○ Using conceptual models and formulating mental models 
○ Judgment and decision making when dealing with complex, incomplete or 
Confidential and Proprietary. Do Not Distribute. 9 
conflicting data 
○ Applying adaptive strategies in rapidly changing situations 
• Human prefers ability to align data to their mental model 
○ Too much data can disconnect human from process 
○ Presentation of data can weaken mental models 
■ Adaptability of display preferences and visualization of data 
■ Human control over decision support tools is key - adapt/accept/reject decisions
Key Considerations for Data Synthesis and 
Visualization 
• New outputs or systems can be ignored or dismissed 
○ Add workload under time constraints 
○ Time and training to learn system may be inadequate 
• Humans must trust and understand system for it to influence “decision” 
• Style of forecaster influences use of tool 
Confidential and Proprietary. Do Not Distribute. 10
Forecasting Data Synthesis Considerations 
• Understand forecasting decision-making better to formulate tools, mitigate 
data overload and ensure optimal visualization/ interaction with systems and 
workstations 
○ Little work has been done to date specific to forecasting community 
○ Differentiate needs between forecasting environments (i.e. severe weather 
forecasting; nominal vs off nominal operations) 
Confidential and Proprietary. Do Not Distribute. 11 
• Heuristics Enhancement 
○ Implement “lessons learned” approach to identify potential biases or possible 
inclination for missed alarm/false alarm 
○ Implement collaborative decision making approach to reduce introduction of 
individual previous experience bias 
• Study best forecasters and poor forecasters to determine what influences 
accuracy and understand performance indicators
References 
Doswell III, D. A. (2004). Weather forecasting by humans - heuristics and decision making. 
Journal of Weather and Forecasting, 19, 1115-1126. 
Gaia, M., & Fontannaz, L. (2006). The human side of weather forecasting. The European 
Confidential and Proprietary. Do Not Distribute. 12 
Forecaster, 14, 17-20. 
Kroonenberg, F. (2000). Human Factor in Severe Weather Forecasting. Paper presented at 9th 
European Conference on Applications of Meteorology/9th European Meteorological 
Society (EMS) Annual Meeting, 28 September – 02 October 2009, Toulouse, France, 
13-10-2009. Retrieved July1, 2014, from 
http://www.emetsoc.org/fileadmin/ems/dokumente/annual_meetings/2009/AM1_EMS20 
09-624.pdf 
Sills, D. M. (2009). On the MSC forecasters forums and the future role of the human forecaster. 
Bulletin of the American Meteorological Society, May 2009, 619-627. 
Stuart, M. A., Schultz, D. M., & Klein, G. (2007). Maintaining the role of humans in the forecast 
process: Analyzing the psyche of expert forecasters. Bulletin of the American 
Meteorological Society, December 2007, 1893-1898.

Optimizing Data Synthesis and Visualization in Real-Time Decision-Making

  • 1.
    Optimizing Data Synthesis and Visualization in Real- Time Decision-Making Kim Bender, AMS Summer Meeting August 2014
  • 2.
    “As weather forecasteryou will be asked to make decisions in situations where you will be under tight time constraints and face uncertainty, you will have huge amounts of data available, some of it contradictory and most of it not really of significance to you, and you will be constantly looking for the correct data, which will not normally be at your disposal (Gaia & Fontannaz, 2006, p.17)” Confidential and Proprietary. Do Not Distribute. 2
  • 3.
    Data Synthesis andVisualization in Real-Time Decision Making • Forecasting and aviation operations are both technological environments where data is increasingly being incorporated into operator workstations to optimize accuracy and efficiency ○ Automation ○ Decision Support Tools ○ Systems Confidential and Proprietary. Do Not Distribute. 3 • Issues and questions: ○ How do you ensure the right systems are incorporated? ○ How do you ensure effective presentation and synthesis of information? ○ What are the impacts of the system on the human operator, safety, and operations? • What predisposes these operations to these environments? Operators are required to make real-time decisions in dynamically changing, uncertain environments
  • 4.
    Human Factors: DecisionMaking in Uncertainty • Decisions Making in Uncertainty ○ Heuristics – intuitive processes based on individual experience and perceptions ○ Objective data – structured and unstructured data but usually statistical in nature • Heuristics mandatory in uncertain environments: Intuition and experience help humans make quick effective decisions ○ Best human forecasters can consistently outperform objective methods- Confidential and Proprietary. Do Not Distribute. 4 Heuristics! ○ Heuristics influenced by past experience and subjective considerations which can be positive or negative ■ Positive – recognize similar situation or pattern and apply lessons learned when data is uncertain ■ Negative – can introduce bias due to previous negative outcome or experience or it can lead to missed event □ Previous false alarm, forecasters tends to under forecast next event leading to a missed alarm □ Previous missed alarm, forecasters tend to over forecast after missed event
  • 5.
    Human Factors: DecisionMaking in Uncertainty • Inordinate amounts of objective data for real-time decision-making can be problematic when not introduced effectively • Operator must mine/culls/filter data to quickly synthesize ○ Multiple data sources can introduce more uncertainty ■ Contradictory output consistencies and forecasts ■ Individual data interpolation and integration/translation differences ○ Data overload – huge amounts of data ■ Contradictory data sources from disparate sources and systems ■ Most data not really of significance for task at hand Confidential and Proprietary. Do Not Distribute. 5 ○ Data Visualization ■ Scanning for correct/relevant data can be time consuming and problematic ■ Relevant data may not always be available or clear ■ Multiple systems introduce display and output inconsistencies If data synthesis and visualization can be structured carefully, could mitigate many significant issues
  • 6.
    Evaluating Effective DataPresentation and Visualization • FAA tests proposed advanced aviation systems, procedures, and concepts and/or performs experimental research to test proposed change ○ Subject Matter Expert (SME) Panels Confidential and Proprietary. Do Not Distribute. 6 ○ Simulation Research ○ Field Research • Introduction of new data, information, and/or display in aviation environment requires: ○ Operational Testing and Evaluation using SME’s ○ Measurement/evaluation of impacts of proposed technologies, displays, or concepts on operations, safety, or human operator ○ SME evaluation of information display
  • 7.
    Key Considerations forData Synthesis and Visualization • Data introduction and visualization considerations ○ Data selection, design, display and is key to performance ○ Only present data that is meaningful and accurate - “too much” data will eventually decrease performance by adding ineffective workload • Implement a systematic approach to data presentation and visualization ○ SME and user evaluations during development and operational testing ○ Disciplined and systematic approach data presentation and visualization display characteristics / system integration ■ Keep systems visually consistent – colors, presentation, warnings, etc. ■ System integration –data consistency Confidential and Proprietary. Do Not Distribute. 7
  • 8.
    Key Considerations forData Synthesis and Visualization • Experimental research to understand the effects of the system on human performance and evaluate alternative approaches • Understand human decision making and how operator visualizes/interacts with large datasets in operational settings ○ Understand each operation – all different ○ Cognitive walkthroughs and walk-through like methods • What data should be presented and how much to automate? ○ Some processes/tasks should not be automated ○ Keep human-in-the-loop to ensure they are active, involved, and aware ○ Research and study need to determine best automation candidates ○ Operational design needs to be structured to tasks Confidential and Proprietary. Do Not Distribute. 8
  • 9.
    Key Considerations forData Synthesis and Visualization • Workstation should make best use of human strengths: ○ Recognizing patterns ○ Using conceptual models and formulating mental models ○ Judgment and decision making when dealing with complex, incomplete or Confidential and Proprietary. Do Not Distribute. 9 conflicting data ○ Applying adaptive strategies in rapidly changing situations • Human prefers ability to align data to their mental model ○ Too much data can disconnect human from process ○ Presentation of data can weaken mental models ■ Adaptability of display preferences and visualization of data ■ Human control over decision support tools is key - adapt/accept/reject decisions
  • 10.
    Key Considerations forData Synthesis and Visualization • New outputs or systems can be ignored or dismissed ○ Add workload under time constraints ○ Time and training to learn system may be inadequate • Humans must trust and understand system for it to influence “decision” • Style of forecaster influences use of tool Confidential and Proprietary. Do Not Distribute. 10
  • 11.
    Forecasting Data SynthesisConsiderations • Understand forecasting decision-making better to formulate tools, mitigate data overload and ensure optimal visualization/ interaction with systems and workstations ○ Little work has been done to date specific to forecasting community ○ Differentiate needs between forecasting environments (i.e. severe weather forecasting; nominal vs off nominal operations) Confidential and Proprietary. Do Not Distribute. 11 • Heuristics Enhancement ○ Implement “lessons learned” approach to identify potential biases or possible inclination for missed alarm/false alarm ○ Implement collaborative decision making approach to reduce introduction of individual previous experience bias • Study best forecasters and poor forecasters to determine what influences accuracy and understand performance indicators
  • 12.
    References Doswell III,D. A. (2004). Weather forecasting by humans - heuristics and decision making. Journal of Weather and Forecasting, 19, 1115-1126. Gaia, M., & Fontannaz, L. (2006). The human side of weather forecasting. The European Confidential and Proprietary. Do Not Distribute. 12 Forecaster, 14, 17-20. Kroonenberg, F. (2000). Human Factor in Severe Weather Forecasting. Paper presented at 9th European Conference on Applications of Meteorology/9th European Meteorological Society (EMS) Annual Meeting, 28 September – 02 October 2009, Toulouse, France, 13-10-2009. Retrieved July1, 2014, from http://www.emetsoc.org/fileadmin/ems/dokumente/annual_meetings/2009/AM1_EMS20 09-624.pdf Sills, D. M. (2009). On the MSC forecasters forums and the future role of the human forecaster. Bulletin of the American Meteorological Society, May 2009, 619-627. Stuart, M. A., Schultz, D. M., & Klein, G. (2007). Maintaining the role of humans in the forecast process: Analyzing the psyche of expert forecasters. Bulletin of the American Meteorological Society, December 2007, 1893-1898.

Editor's Notes

  • #9 Maintain skills and keep professional “fresh” Help professional to identify faulty or suspicious data and question system when necessary Aid situation awareness, memory, or formulate mental models
  • #11 Big data is an all-encompassing term for any collection of data sets so large and complex that it becomes difficult to process using on-hand data management tools or traditional data processing applications. Style of forecaster – some find creative and positive ways of using new data and visualizations others are resistant to change (education and training crucial)