nickpowdthavee.bsky.social
@nickpowdthavee.bsky.social
Full-time happiness economist. Part-time behavioural economist, working on the edge of AI. Professor of Economics at Nanyang Technological University. FREcon.
Bottom line: LLMs are promising diagnostic tools, but they are no substitute for listening to people talk about how they feel about their lives.
๐ Open access in @PNASNews
โSimulating human well-being with large language models.โ
www.pnas.org/doi/pdf/10.1...
๐ Open access in @PNASNews
โSimulating human well-being with large language models.โ
www.pnas.org/doi/pdf/10.1...
PNAS
Proceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...
www.pnas.org
December 9, 2025 at 4:18 PM
Bottom line: LLMs are promising diagnostic tools, but they are no substitute for listening to people talk about how they feel about their lives.
๐ Open access in @PNASNews
โSimulating human well-being with large language models.โ
www.pnas.org/doi/pdf/10.1...
๐ Open access in @PNASNews
โSimulating human well-being with large language models.โ
www.pnas.org/doi/pdf/10.1...
๐ค They reproduce broad patterns (richer, healthier โ more satisfied with life)
โ ๏ธ But systematically misestimate well-being
๐ Especially in countries least represented online
๐ LLMs rely mostly on linguistic shortcuts, not lived experience
โ ๏ธ But systematically misestimate well-being
๐ Especially in countries least represented online
๐ LLMs rely mostly on linguistic shortcuts, not lived experience
PNAS
Proceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...
www.pnas.org
December 9, 2025 at 4:18 PM
๐ค They reproduce broad patterns (richer, healthier โ more satisfied with life)
โ ๏ธ But systematically misestimate well-being
๐ Especially in countries least represented online
๐ LLMs rely mostly on linguistic shortcuts, not lived experience
โ ๏ธ But systematically misestimate well-being
๐ Especially in countries least represented online
๐ LLMs rely mostly on linguistic shortcuts, not lived experience
Can AI really predict how satisfied we are with our lives? We tested 4 leading language models on 64,000 people across 64 countries, and found:
PNAS
Proceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...
www.pnas.org
December 9, 2025 at 4:18 PM
Can AI really predict how satisfied we are with our lives? We tested 4 leading language models on 64,000 people across 64 countries, and found: