Better Data for Better Policy: Opportunities and Challenges
- 1. Stefaan G. Verhulst OPPORTUNITIES AND CHALLENGES BETTER DATA FOR BETTER POLICY
- 2. Premise 1 Data can Inform and Transform the full Policy Life Cycle in 4 Ways
- 4. 4 WAYS DATA CAN INFORM/TRANSFORM THE POLICY CYCLE
- 5. SITUATION ANALYSIS DATACANINFORMANDTRANSFORMTHEFULLPOLICYLIFECYCLEIN4WAYS PREMISE1 EVOLVING POLICY CICLE P RIORITAZION SOLUTION ITERATION IM PLEMENTATIONREVIEW AGENDASETTING PROB LEM DEFINITION DESIGN EN FORCEMENTEVALUATION
- 6. DATACANINFORMANDTRANSFORMTHEFULLPOLICYLIFECYCLEIN4WAYS PREMISE1 SITUATION ANALYSIS EVOLVING POLICY CICLE P RIORITAZION SOLUTION ITERATION IM PLEMENTATIONREVIEW AGENDASETTING PROB LEM DEFINITION DESIGN EN FORCEMENTEVALUATION
- 7. DATACANINFORMANDTRANSFORMTHEFULLPOLICYLIFECYCLEIN4WAYS PREMISE1 KNOWLEDGE CREATION EVOLVING POLICY CICLE P RIORITAZION SOLUTION ITERATION IM PLEMENTATIONREVIEW AGENDASETTING PROB LEM DEFINITION DESIGN EN FORCEMENTEVALUATION
- 8. DATACANINFORMANDTRANSFORMTHEFULLPOLICYLIFECYCLEIN4WAYS PREMISE1 KNOWLEDGE CREATION/TRANSFER EVOLVING POLICY CICLE P RIORITAZION SOLUTION ITERATION IM PLEMENTATIONREVIEW AGENDASETTING PROB LEM DEFINITION DESIGN EN FORCEMENTEVALUATION
- 11. DATACANINFORMANDTRANSFORMTHEFULLPOLICYLIFECYCLEIN4WAYS PREMISE1 EVALUATION AND IMPACT ASSESSMENT EVOLVING POLICY CICLE P RIORITAZION SOLUTION ITERATION IM PLEMENTATIONREVIEW AGENDASETTING PROB LEM DEFINITION DESIGN EN FORCEMENTEVALUATION
- 12. DATACANINFORMANDTRANSFORMTHEFULLPOLICYLIFECYCLEIN4WAYS PREMISE1 EVALUATION AND IMPACT ASSESSMENT EVOLVING POLICY CICLE P RIORITAZION SOLUTION ITERATION IM PLEMENTATIONREVIEW AGENDASETTING PROB LEM DEFINITION DESIGN EN FORCEMENTEVALUATION
- 14. AGILE AND DATA DRIVEN POLICY MAKING Requires Data Sets that are TIME-SENSITIVEFINE-GRAINED DATACANINFORMANDTRANSFORMTHEFULLPOLICYLIFECYCLEIN4WAYS PREMISE1
- 15. Premise 2 New Data Sources Can Make the Policy Cycle More Agile and Informed
- 21. PRIVATE DATA NEW DATA SOURCES Web Crawling Web Scraping Web Search Analysis Commercial Transactions Scanner Data Credit Card Data Sensor and Geospatial Data Social Media Telecom Data
- 22. ADVANCES IN AI, DATA AND COMPUTING SCIENCE SENTIMENT ANALYSIS SCRAPING CDR ANALYTICS NATURAL LANGUAGE PROCESSING DATA MINING MACHINE LEARNING PREDICTIVE ANALYTICS ETC.
- 23. Premise 3 Access to New Data Sources Requires new Partnerships: Data Collaboratives
- 24. DATA COOPERATIVES OR POOLING PRIZES & CHALLENGES RESEARCH PARTNERSHIPS INTELLIGENCE PRODUCTS TRUSTED INTERMEDIARY APPLICATION PROGRAMMING INTERFACES (APIS) SIX TYPES OF DATA COLLABORATIVES
- 25. DATA POOLING
- 29. API
- 31. MOTIVATIONS TO SHARE: THE SIX Rs BEHIND CORPORATE DATA SHARING RESPONSIBILITY REGULATORY COMPLIANCE REVENUE REPUTATION & RETAINMENT OF TALENT RECIPROCITY RESEARCH & INSIGHTS
- 32. BUT WE CAN DO BETTER THAN VOTING.
- 35. PRIVACY & SECURITY COMP ETITIVECONCERNS GENERALIZABILITY & DATAQUALITYCULTURALCHALL ENGES CONCERNS
- 36. RIS K AND VALUE ASSESSME NT METHO DSANDTECHNOLOGIES PRINCIPLESANDPROC ESSES TOWARD DATA RESPONSIBILITY
- 37. COLLECTION PROCESSING SHARING ANALYZING USING Collecting inaccurate, old or “dirty” data affecting data quality and ability draw meaningful insights Unauthorized or intrusive data collection potentially leading to privacy harms; Lack of interoperable cultural and institutional norms and expectations, creating a difficult environment to collaborate toward mutual benefit; Lack of data stewardship at both ends to ensure the responsible use of personally identifiable information as it travels across use cases and sectors; Poor problem definition or research design, potentially leading to data being analyzed in a way that does not add value toward the ultimate objective When data is ultimately put to use, risks emerge especially from collaborative organizations using shared data controversially or incongruously in relation to the original objective for its collection and/ or the original consent provided by the data subject (if any). Such risks can have negative results like the misinterpretation of data, the re-identification of individual data subjects, and decisional interference Insufficient, outdated or inflexible security provisions creating the potential for data vulnerabilities or breaches; Incomplete or non-representative sampling of the universe – e.g., ignoring “data invisibles,” or population segments with a limited data footprint – potentially leading to non-inclusive or unrepresentative approaches or interventions. Improper or unauthorized access to shared data as it passes between entities, whether by unsanctioned actors inside or outside of collaborating organizations; Conflicting legal jurisdictions and different levels of security within collaborating entities, making the eventual congruous data use difficult. Inaccurate data modeling or use the of biased algorithms, which, like dirty data at the Collection stage, can lead to confidence in fundamentally flawed insights. Additionally, at the Using stage, many of the risks from previous stages could yield true, identifiable harms for the first time – e.g., a negatively impactful policy decision being made based on faulty data from the Collection stage. Aggregation and correlation of incompatible datasets can create ‘apples and oranges’ scenarios where the eventual sharing and analysis of commingled datasets are doomed for failure. TOWARD DATA RESPONSIBILITY VALUE AND RISK ASSESSMENT
- 38. UPDATING FIPPS/IRB? THE SIGNAL CODE The Signal Code articulates five human rights to information during crisis: THE RIGHT TO INFORMATION THE RIGHT TO PROTECTION FROM HARM THE RIGHT TO DATA SECURITY AND PRIVACY THE RIGHT TO DATA AGENCY THE RIGHT TO REDRESS AND RECTIFICATION Building Robust Review for Industry Research Molly Jackman Facebook Lauri Kanerva Facebook EVOLVING THE IRB TOWARD DATA RESPONSIBILITY PRINCIPLES AND PROCESSES
- 39. Differential Privacy? New Methods and Technologies Decision Trees TOWARD DATA RESPONSIBILITY METHODS AND TECHNOLOGIES
- 40. METHODS MOVEMENTEVIDENCE STEWARDS REALIZING THE 3 PREMISES TOWARD DATA DRIVEN POLICY MAKING