Ham Huang
500 posts
@Huang_Ham
PhD student @Princeton Psych under Drs. Natalia Vélez & Tom Griffiths, studying the computational cognition of human aggregate minds. Before @Penn @Cal
🦔Researchers at the University of Pennsylvania studied what they call cognitive surrender, the tendency to accept AI outputs without critical evaluation. Across 1,372 participants and over 9,500 trials, subjects accepted faulty AI reasoning 73.2% of the time and only overruled it 19.7% of the time. When the AI was wrong, users still accepted its answer 80% of the time. Subjects who used AI scored 11.7% higher on confidence in their answers despite the AI being wrong half the time. Adding time pressure made people 12 percentage points less likely to catch AI errors. Adding financial incentives and immediate feedback made them 19 points more likely to catch them. My Take The time pressure finding matters enormously for how AI is actually being deployed in workplaces. Companies are using AI to justify faster turnaround times, which means employees are using it under exactly the conditions that make them least likely to catch mistakes. When you're rushed, your internal monitor for detecting errors essentially stops firing, so you get AI output, no time to review it, high confidence it's correct, and a meaningful chance it's wrong. People using a system that was wrong half the time still felt more confident in their answers than people who weren't using AI at all. That is a system actively making people worse at knowing what they don't know, which is one of the most dangerous things you can do to human judgment at scale. The companies pushing AI hardest into employee workflows should be reading this research carefully. Hedgie🤗 Link to research for those interested: papers.ssrn.com/sol3/papers.cf…
Humans can use positive and negative spectrotemporal correlations to detect rising and falling pitch dlvr.it/TQs6Nm
LLMs develop novel biases from experience. New preprint: LLMs that make decisions & get feedback develop new views — including ⚠️harmful stereotypes that target demographics! [1/7]
Now out in Nature Human Behaviour! 🚀🚀 Over the past decades, research on collective human behaviour has relied heavily on networks. This is intuitive: people interact with other people. However, we argue that this dominant framework misses a crucial ingredient. Traditional networks represent agents as nodes and pairwise relations as edges. As a result, they fundamentally assume that social interactions can be decomposed into pairs. Yet many social processes are irreducibly group-based. A simple example: a group of three coauthors writing a paper cannot be reduced to three independent pairs of coauthors. The group itself matters. In this article, we review a wide range of empirical and theoretical cases where group interactions cannot be decomposed into pairwise ones, and show that higher-order interactions shape collective behaviour above and beyond dyadic ties. We advocate studying collective behaviour on hypergraphs, where interactions can involve multiple agents simultaneously. We review how hypergraphs provide new insights across domains, including affiliation and collaboration networks, high-frequency contact settings (families, friends), and key social processes such as social contagion, cooperation, truth-telling, and moral behaviour. Finally, we outline promising directions for future research: addressing computational challenges of higher-order models; studying bias and inequality in group dynamics; combining hypergraphs and large language models to investigate the coevolution of language and behaviour; and using higher-order networks to simulate the impact of policies before implementation; and others. We are very excited about this work and hope it will inspire further research in a rapidly growing and fundamental area with broad real-world implications. Link to the paper in the first reply This work was brilliantly led by Federico Battiston (@fede7j), with an outstanding team of co-authors: Fariba Karimi (@fariba_k), Sune Lehmann, Andrea Bamberg Migliano, Onkar Sadekar (@OnkarSadekar), Angel Sanchez, & Matjaz Perc (@matjazperc)
