Sign up to save your podcasts
Or
Share Making Machines Understand Humans
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
Making Machines Understand Humans
Making Machines Understand Humans: Behavioral Data Science, Trust, and AI’s Real Risks
Data scientist and behavioral decision theorist Ganna Pogrebna explains behavioral data science as making machines understand human preferences, goals, and values, outlining three strands: economic psychology, mathematical analysis of algorithm behavior, and engineering of complex human–machine systems.
She discusses ethical concerns around “trust” in probabilistic large language models, how algorithmic recommendations shape user agency, and the uncertain impacts of relationship AI amid rising loneliness and changing communication. Pogrebna argues AI can both degrade and enhance skills depending on use, citing evidence on navigation and a radiology study where weaker doctors worsened while stronger improved with decision support.
She emphasizes LLMs as pattern-imitating creativity tools prone to errors, warns of “data inbreeding” as human data becomes scarce, and advocates defining problems first—often solvable via behavioral design—plus diverse oversight to anticipate harms. She advises young people to build transferable skills, be AI-literate, and pursue real-world problems, and highlights her future focus on data-scarce modeling and quantum decision approaches.
00:00 Meet Ganna Pogrebna
01:01 What Behavioral Data Science Does
02:15 Three Strands of the Field
04:35 Trust Ethics and Agency
07:21 Relationship AI and Loneliness
11:53 Skills We Lose and Gain
14:40 System One vs System Two
17:45 LLMs as Creativity Tools
22:02 Illusion of Understanding
24:03 AGI Consciousness Skepticism
27:26 Data Inbreeding and Human Data
30:25 AI Homogenizes Our Voice
31:40 AI for Smarter Cities
32:48 Cities Need Human Focus
33:29 Time Perception Happiness
34:58 Designing Time Stoppers
37:15 Behavioral Design Wins
39:06 Do You Need AI
42:52 AI Risk And Context
47:01 Anticipating Harms Early
49:54 Government Automation Paths
53:59 Diversity Beyond Optics
01:00:13 Advice For Teen Learners
01:05:47 Quantum Decision Making
01:08:22 Closing Thanks
https://www.gannapogrebna.com/
View all episodes
By Andrew PhippsMaking Machines Understand Humans: Behavioral Data Science, Trust, and AI’s Real Risks
Data scientist and behavioral decision theorist Ganna Pogrebna explains behavioral data science as making machines understand human preferences, goals, and values, outlining three strands: economic psychology, mathematical analysis of algorithm behavior, and engineering of complex human–machine systems.
She discusses ethical concerns around “trust” in probabilistic large language models, how algorithmic recommendations shape user agency, and the uncertain impacts of relationship AI amid rising loneliness and changing communication. Pogrebna argues AI can both degrade and enhance skills depending on use, citing evidence on navigation and a radiology study where weaker doctors worsened while stronger improved with decision support.
She emphasizes LLMs as pattern-imitating creativity tools prone to errors, warns of “data inbreeding” as human data becomes scarce, and advocates defining problems first—often solvable via behavioral design—plus diverse oversight to anticipate harms. She advises young people to build transferable skills, be AI-literate, and pursue real-world problems, and highlights her future focus on data-scarce modeling and quantum decision approaches.
00:00 Meet Ganna Pogrebna
01:01 What Behavioral Data Science Does
02:15 Three Strands of the Field
04:35 Trust Ethics and Agency
07:21 Relationship AI and Loneliness
11:53 Skills We Lose and Gain
14:40 System One vs System Two
17:45 LLMs as Creativity Tools
22:02 Illusion of Understanding
24:03 AGI Consciousness Skepticism
27:26 Data Inbreeding and Human Data
30:25 AI Homogenizes Our Voice
31:40 AI for Smarter Cities
32:48 Cities Need Human Focus
33:29 Time Perception Happiness
34:58 Designing Time Stoppers
37:15 Behavioral Design Wins
39:06 Do You Need AI
42:52 AI Risk And Context
47:01 Anticipating Harms Early
49:54 Government Automation Paths
53:59 Diversity Beyond Optics
01:00:13 Advice For Teen Learners
01:05:47 Quantum Decision Making
01:08:22 Closing Thanks
https://www.gannapogrebna.com/