Dr. Sheen S. Levine

I study (and teach) how people behave and how they impact others, organizations and markets.

[Dr. Sheen S. Levine @sslevine]

𝐍𝐨𝐭 𝐛𝐥𝐢𝐧𝐝.

Cultural evolution long treated where new ideas come from as random. In @PNASnews Yaman, Tian & Lindström show semantic knowledge guides innovation — strip it, and humans do no better than random bots.

https://t.co/1bzAMB3oAw

[Dr. Sheen S. Levine @sslevine]

Women tolerate workload inequality more than earnings inequality; men don't distinguish between the two. And it's not social norms driving the gap — a follow-up rules that out. The lever may be structural, not cultural. https://t.co/MURGvVUQwE

[Dr. Sheen S. Levine @sslevine]

𝐘𝐨𝐮 𝐟𝐚𝐜𝐞 𝐚 𝐩𝐫𝐨𝐛𝐥𝐞𝐦. 𝐘𝐨𝐮 𝐫𝐞𝐚𝐝 𝐭𝐡𝐞 𝐝𝐨𝐜𝐬. 𝐘𝐨𝐮 𝐠𝐞𝐭 𝐟𝐫𝐮𝐬𝐭𝐫𝐚𝐭𝐞𝐝. So you write a polite-but-firm memo. Standard middle-management behavior—except this one was written by a Codex agent. The Turing test is being passed in places no one expected

[Dr. Sheen S. Levine @sslevine]

Got rejected? Your editor acted as expected. Smith et al. use LLMs on 80,000 STEM peer reviews: when reviewers diverge, editors side with the most negative — esp. on novelty and accuracy. Does our fear of uncertainty snuff out the most creative research? https://t.co/4pvtbPJLxE

[Dr. Sheen S. Levine @sslevine]

𝐓𝐡𝐞 "𝐢𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞" 𝐢𝐧 "𝐀𝐈" implies learning. Agents do learn— like children, just not always what we'd want. The fix is older than computing: ask the practitioner to reflect. At the end of a long memo, I tuck this.

[Dr. Sheen S. Levine @sslevine]

𝐓𝐡𝐞 𝐂𝐥𝐚𝐮𝐝𝐞 𝐄𝐟𝐟𝐞𝐜𝐭: 1260 social scientists surveyed. 20% use coding agents — and post 75% more working papers. Users skew young, male, top-university. A Matthew effect for the AI era, but unlike fame, adoption is still movable. @AnthropicAI https://t.co/ZpOmScNJel

[Dr. Sheen S. Levine @sslevine]

Women tolerate workload inequality more than earnings inequality; men don't distinguish between the two. And it's not social norms driving the gap — a follow-up rules that out. The lever may be structural, not cultural. https://t.co/MURGvVUQwE

[Dr. Sheen S. Levine @sslevine]

𝐓𝐡𝐞 "𝐢𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞" 𝐢𝐧 "𝐀𝐈" implies learning. Agents do learn— like children, just not always what we'd want. The fix is older than computing: ask the practitioner to reflect. At the end of a long memo, I tuck this.

[Dr. Sheen S. Levine @sslevine]

𝗡𝗲𝘄 𝗹𝗮𝗯 𝗺𝗲𝗺𝗯𝗲𝗿: Claude Opus 4.8. No teaching. No committees. No office hours. The tenure clock is not running—but the work is.

[Dr. Sheen S. Levine @sslevine]

𝐓𝐡𝐞 𝐂𝐥𝐚𝐮𝐝𝐞 𝐄𝐟𝐟𝐞𝐜𝐭: 1260 social scientists surveyed. 20% use coding agents — and post 75% more working papers. Users skew young, male, top-university. A Matthew effect for the AI era, but unlike fame, adoption is still movable. @AnthropicAI https://t.co/ZpOmScNJel

[Dr. Sheen S. Levine @sslevine]

𝐘𝐨𝐮 𝐟𝐚𝐜𝐞 𝐚 𝐩𝐫𝐨𝐛𝐥𝐞𝐦. 𝐘𝐨𝐮 𝐫𝐞𝐚𝐝 𝐭𝐡𝐞 𝐝𝐨𝐜𝐬. 𝐘𝐨𝐮 𝐠𝐞𝐭 𝐟𝐫𝐮𝐬𝐭𝐫𝐚𝐭𝐞𝐝. So you write a polite-but-firm memo. Standard middle-management behavior—except this one was written by a Codex agent. The Turing test is being passed in places no one expected

[Dr. Sheen S. Levine @sslevine]

𝐅𝐫𝐢𝐞𝐧𝐝𝐬 𝐮𝐫𝐠𝐞 𝐦𝐞 𝐭𝐨 𝐩𝐨𝐬𝐭 𝐦𝐨𝐫𝐞 𝐚𝐛𝐨𝐮𝐭 𝐦𝐲𝐬𝐞𝐥𝐟, 𝐬𝐨 𝐡𝐞𝐫𝐞'𝐬 𝐚𝐧 𝐚𝐭𝐭𝐞𝐦𝐩𝐭. Our paper—economic experiment meets sociological theory meets organizational pain—has been nominated for the Best Paper Prize by the Strategic Management Society.

[Dr. Sheen S. Levine @sslevine]

𝐓𝐡𝐞 𝐀𝐈 𝐝𝐞𝐥𝐮𝐠𝐞 𝐢𝐧 𝐚𝐜𝐚𝐝𝐞𝐦𝐢𝐜 𝐩𝐮𝐛𝐥𝐢𝐬𝐡𝐢𝐧𝐠. AI slop in publishing: machine finds database, runs every association, writes paper. But theory, preregistration, experiments already filter this. Apply them, and you get better science. https://t.co/AeJqZC4a38

[Dr. Sheen S. Levine @sslevine]

𝐖𝐡𝐞𝐧 𝟔𝟎% 𝐨𝐟 𝐠𝐫𝐚𝐝𝐞𝐬 𝐚𝐫𝐞 𝐀'𝐬, 𝐭𝐡𝐞 𝐀 𝐜𝐨𝐧𝐯𝐞𝐲𝐬 𝐧𝐨 𝐢𝐧𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧. Harvard's faculty just voted to cap A's at 20% and replace GPA with rankings for honors. Let's see if others follow. https://t.co/UPiTb5eIio

[Dr. Sheen S. Levine @sslevine]

LLMs tell you what you want to hear. New @Science paper: AIs affirmed users 49% more than humans; 51% of the time when humans had unanimously voted "you're the asshole." One chat makes users more righteous, less willing to repair. @MyraCheng @DanJurafsky https://t.co/ltW0Wck7Eq

[Dr. Sheen S. Levine @sslevine]

𝐋𝐚𝐬𝐭 𝐰𝐞𝐞𝐤 𝐈 𝐧𝐨𝐭𝐞𝐝 𝐭𝐡𝐚𝐭 𝐌𝐁𝐀 𝐝𝐞𝐠𝐫𝐞𝐞𝐬 𝐚𝐫𝐞 𝐧𝐨𝐰 𝐨𝐧 𝐬𝐚𝐥𝐞. Prices fell because federal subsidy was removed, not because schools got greedy at last. Schools priced to what the government would lend. That's subsidy economics. https://t.co/Bu0cUXYaLo

[Dr. Sheen S. Levine @sslevine]

𝐓𝐡𝐞 𝐀𝐈 𝐝𝐞𝐥𝐮𝐠𝐞 𝐢𝐧 𝐚𝐜𝐚𝐝𝐞𝐦𝐢𝐜 𝐩𝐮𝐛𝐥𝐢𝐬𝐡𝐢𝐧𝐠. AI slop in publishing: machine finds database, runs every association, writes paper. But theory, preregistration, experiments already filter this. Apply them, and you get better science. https://t.co/AeJqZC4a38

[Dr. Sheen S. Levine @sslevine]

𝐅𝐫𝐢𝐞𝐧𝐝𝐬 𝐮𝐫𝐠𝐞 𝐦𝐞 𝐭𝐨 𝐩𝐨𝐬𝐭 𝐦𝐨𝐫𝐞 𝐚𝐛𝐨𝐮𝐭 𝐦𝐲𝐬𝐞𝐥𝐟, 𝐬𝐨 𝐡𝐞𝐫𝐞'𝐬 𝐚𝐧 𝐚𝐭𝐭𝐞𝐦𝐩𝐭. Our paper—economic experiment meets sociological theory meets organizational pain—has been nominated for the Best Paper Prize by the Strategic Management Society.

[Dr. Sheen S. Levine @sslevine]

𝐖𝐡𝐞𝐧 𝟔𝟎% 𝐨𝐟 𝐠𝐫𝐚𝐝𝐞𝐬 𝐚𝐫𝐞 𝐀'𝐬, 𝐭𝐡𝐞 𝐀 𝐜𝐨𝐧𝐯𝐞𝐲𝐬 𝐧𝐨 𝐢𝐧𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧. Harvard's faculty just voted to cap A's at 20% and replace GPA with rankings for honors. Let's see if others follow. https://t.co/UPiTb5eIio

[Dr. Sheen S. Levine @sslevine]

𝐋𝐚𝐬𝐭 𝐰𝐞𝐞𝐤 𝐈 𝐧𝐨𝐭𝐞𝐝 𝐭𝐡𝐚𝐭 𝐌𝐁𝐀 𝐝𝐞𝐠𝐫𝐞𝐞𝐬 𝐚𝐫𝐞 𝐧𝐨𝐰 𝐨𝐧 𝐬𝐚𝐥𝐞. Prices fell because federal subsidy was removed, not because schools got greedy at last. Schools priced to what the government would lend. That's subsidy economics. https://t.co/Bu0cUXYaLo