Interpreting Data: Wielding Data for Good and Evil in the AI Era

Data is a powerful tool that permeates our everyday lives. As both users of data systems and market consumers, we know how data can be a superpower for good and evil. Increasing technology capabilities bring more opportunities for benefit and destruction, further expanding in today's AI era. How can we balance these seemingly competing interests? And are the benefits worth the cost?
In this session, we will explore ways data can be used to make the world a better place or be used against us. We will discuss technologies and tools that help us as developers wield it for good and recognize and combat misuse. Learn how to use data to make a positive impact on the world and set up safe practices for the future.
Demo code: https://github.com/JMHReif/graph-demo-datasets/tree/main/data-good-evil

Transcript

1. Interpreting Data: Wielding Data for Good and Evil in the AI Era Jennifer Reif jennifer.reif@neo4j.com @JMHReif github.com/JMHReif jmhreif.com linkedin.com/in/jmhreif
2. Who is Jennifer Reif? Developer Advocate, Neo4j • Continuous learner • Technical speaker • Tech blogger, podcaster • Other: geek Jennifer Reif jennifer.reif@neo4j.com @JMHReif github.com/JMHReif jmhreif.com linkedin.com/in/jmhreif
3. Person of Interest (tv series) 2011-2016 • Programmer built an AI for government to predict terrorist threats • Gov ignores “irrelevant” threats • Protect victims / bring justice to perpetrators • Raises moral issues, including… • Privacy • Greater good • AI + security
4. Data is common, yet powerful
5. Issue #1: Value of data • Tangible vs intangible assets • Time-contingent value • Transaction vs health data • Avoid deleting data…might need it later • Variable value and strategies • Cultures / countries • Individual risks vs rewards
6. Issue #2: Perceived trust • Take systems at their word • Systems built on math and facts • Humans make mistakes • Except…GenAI • Probabilities and human data • Error-prone
7. Data as consumers Make decisions • Weather forecasts -> what to wear • Nutrition content -> what to eat • Product reviews/research -> what to buy • Forums/blogs -> what to learn • Newsletters, social/community -> what’s going on • Services, fi nances, etc -> what to do
8. Systems we build… Do we consider the results and potential? • How will data be used? • Downstream impacts? • Data quality? • Solution accuracy? • Results over time?
9. Data for Good and Evil • Bene fi ts: • Personalization • Convenience • Life improvement • Risks: • Security (hacks / leaks) • Desensitization • Maintenance • Digital Twin of person
10. In a world of AI… Accelerating + Amplifying • Morality / ethics • Data quality • Humanity • Transparency • Evaluation
11. Part 1: Data for Evil
12. Person of Interest Surveillance, Control, Manipulation • Government • Irrelevant threats • Private • Best intentions morph over time • Pandora’s Box… • Once opened, can’t close
13. Misuse of data What can happen? • Private data -> public • Data pro fi t without consent • Manipulation, misrepresentation • Ruined fi nances, respect, life
14. What happens if data is simply wrong? Quality matters too • Name misspell or truncation • Di ffi cult to social friend or contact • Partial match conviction • Data duplication / falsi fi cation • Inaccurate aggregations / stats • Identity theft, fraud, etc
15. Missing data / gaps One example • Maternity jeans • Stretch / comfort • Supportive in changes • Useful in other uses… • Post-surgery recovery • Internal illnesses • Caution: limiting applicability and searchability!
16. Consequences Person of Interest • Unchecked power • Changing intentions • Higher and higher stakes • Unknown impacts!
17. Part 2: Data for Good
18. Person of Interest Predict and protect • Privacy: provide number as ID • No one is irrelevant • Predict threats, save lives • Human intervention for details + action
19. Light side of AI and data • Healthcare: • Algorithms for diagnosing diseases • Optimize treatment plans • Nature: • Understand patterns • Monitor changes • Education: • Personalized learning • Maximize value for diverse needs
20. Data for empowerment and progress • Enabler • Human limitations can be improved or overcome • Access • Opens information to broader audience • Participation • Opens doors for involvement • Solutions • Solve problems couldn’t solve before
21. Hyper-Personalization One small example / opportunity • Personalization can lead to cyclic results • Some randomness needed • Depends on preferences, mood, etc • Customization option: • Dial up/down randomness • Interactive personalization by user
22. Demo!
23. Striking Balance - Good vs Evil
24. Duality of Developers Creators and consumers • Developers build the systems • Can help shape outcomes • Can consider / evaluate • Design • Data (minimum collected) • Use • Evolution over time
25. Duality of Developers Insider knowledge • AI tools • Data processes / transformations • Awareness of design and fl aws • Use to inform design, development, and rules • Transparency + quality is key
26. Tools and Tech for Safe Practices • Weigh pros and cons • Safeguards: • Evaluate impacts + risks • Establish practices and tech • Data: • Transparent, accountable, responsible • Practices and tech
27. Developers as consumers Advocates • Evaluate • Diversify • Research -> cautious trust • Humans / Companies / Systems • Re-evaluate • Opt-out
28. Developers as builders Advocates • Insight into use and value • Advocate for… • Better guardrails • Better design / testing / response • Transparency • Build trust and participation for better results
29. Road Ahead: Ensuring Good
30. Emerging trends already Prevent bad, assist good • AI explainability, fairness, transparency • Human-centered design • Gate-checks, rules, human intervention steps • Intellectual property • Societal / human impacts
31. Developers shape the future • Humanity 1st • Data quality • Identify problems -> solutions • Participate and document decisions • Iterate and improve
32. Tradeoff: cost-benefit analysis Depends on preferences Consolidate vs. Diversify Convenience vs. Complexity
33. - Harold Finch, “2-Pi-R” Pi…it keeps on going, forever, without ever repeating. Which means that contained within this string of decimals, is every single other number…And if you convert these decimals into letters, you would have every word that ever existed… Now what you do with that information; what it’s good for, well that would be up to you. Jennifer Reif jennifer.reif@neo4j.com @JMHReif github.com/JMHReif jmhreif.com linkedin.com/in/jmhreif Demo: github.com/JMHReif/graph-demo-datasets/data-good-evil