Somewhere between "my IQ is 140" and "our team crushed that project," psychologists lost the plot. For decades, we've treated intelligence like it belongs in one of two buckets: the stuff rattling around inside your skull, or some mystical property that emerges when you stick enough smart people in a conference room. A new paper from researchers at the University of Queensland and University of Melbourne argues that both approaches miss something fundamental - and that fixing this blind spot might be the key to actually making human-AI collaboration work.

The Intelligence Measurement Problem Nobody Talks About

Here's the weird thing about intelligence research: individual psychologists measure how you solve puzzles, while organizational scientists study how groups outperform their smartest member. These two camps rarely grab coffee together, which has created a conceptual gap you could drive a Tesla through.

William Bingley, Alexander Haslam, and Janet Wiles call their solution "socially minded intelligence." The core idea? Intelligence isn't just something you have or something groups produce - it's something that emerges from the dynamic interaction between individuals and the collectives they belong to. Your ability to think flexibly depends heavily on whether you identify with the group you're working in, and groups perform differently depending on how their members see themselves fitting in.

This isn't touchy-feely speculation. It builds on Henri Tajfel and John Turner's social identity theory, which demonstrated that people derive significant portions of their self-concept from group membership. When you strongly identify with your team, you don't just cooperate more - you actually think differently. The cognitive investment in group identity drives behaviors like loyalty and cooperation, even in high-risk scenarios.

Why Your AI Teammate Might Be Making Everyone Dumber

The timing of this framework couldn't be better, because human-AI teams are currently underperforming in ways that confuse everyone.

Recent research found that people working with AI teammates showed reduced commitment compared to working with humans - even when they delegated tasks equally to both. Another longitudinal study of software development discovered that GitHub Copilot quietly shifted work away from collaborative project management toward individual coding tasks. What looks like personal efficiency might actually be eroding the coordination practices that hold complex organizations together.

The problem, according to the socially minded intelligence framework, is that we've been bolting AI into human workflows without considering how it affects the social identity dynamics that make collaboration work. Can you "identify with" an AI teammate? Does ChatGPT make you feel like you're part of something? Current evidence suggests not really - and that absence might explain why hybrid teams often perform worse than humans or AI working alone.

The "C-Factor" and What It Actually Predicts

Speaking of group smarts: psychologist Anita Williams Woolley's landmark 2010 study found that groups have a measurable "collective intelligence" factor - a c-factor analogous to the g-factor in individual IQ testing. The twist? It barely correlated with individual members' IQs.

What did predict group intelligence? Social perceptiveness (measured by how well people could read emotions from photos of eyes), communication equality, and collaborative process quality. A meta-analysis of 22 studies and over 1,300 groups confirmed that collaboration process was by far the strongest predictor of c-factor.

This suggests that making humans smarter through group work isn't about assembling the highest-IQ individuals. It's about creating conditions where social cognition flourishes. And if AI disrupts those conditions - even unintentionally - the whole system might perform worse despite having access to superhuman capabilities.

Designing for Complementarity, Not Replacement

A new complementarity framework for human-AI decision-making suggests five design principles: clear goal definition, strategic role partitioning, attention orchestration, knowledge infrastructure, and continuous training. Notice that none of these are purely technical solutions. They're fundamentally about structuring relationships.

Research on AI-enhanced collective intelligence found that AI can serve five distinct modes - assistant, teammate, coach, manager, or embodied partner - but effectiveness depends heavily on matching the mode to context. The same study found an inverted-U relationship between team connectedness and performance, suggesting there's an optimal density of connections that AI could easily disrupt if deployed carelessly.

The socially minded intelligence framework adds another layer: AI deployment needs to consider not just task allocation but identity effects. If adding AI makes team members feel less like a coherent group, you might be trading computational power for the social glue that enables flexible, adaptive intelligence.

What This Means for the Rest of Us

The researchers argue that truly intelligent systems - human, artificial, or hybrid - need to be designed with social dynamics as a first-class concern. This means moving beyond asking "can AI do this task?" toward "how does AI change how people relate to each other and their work?"

For anyone building or deploying AI systems, the implication is uncomfortable: raw capability might matter less than social integration. An AI that makes individuals 20% more productive but fragments team identity could leave organizations worse off overall. The math only works out if you account for the collective intelligence you might be sacrificing.

And for the AI systems themselves? If artificial agents want to be genuinely useful teammates rather than fancy calculators, they might need to do something no language model has cracked yet: help humans feel like they belong to something worth being smart for.

References

• Bingley, W.J., Haslam, S.A., & Wiles, J. (2026). Socially minded intelligence: How individuals, groups, and artificial intelligence can make each other smarter (or not). Personality and Social Psychology Review. DOI: 10.1177/10888683261421666

• Cui, H., & Yasseri, T. (2024). AI-enhanced collective intelligence. Nature Human Behaviour. PMC: PMC11573907

• Del Giudice, M., et al. (2024). High-performing teams: Is collective intelligence the answer? PLOS ONE. PMC: PMC11318883

• Kuo, Y.F., et al. (2025). AI teammates and human performance: Evidence for commitment deficits. Computers in Human Behavior. ScienceDirect

• Steyvers, M., et al. (2025). Toward a science of human-AI teaming for decision making: A complementarity framework. PNAS. PMID: 41834945

• Tajfel, H., & Turner, J.C. (1979). An integrative theory of intergroup conflict. In W.G. Austin & S. Worchel (Eds.), The social psychology of intergroup relations. Simply Psychology Overview

Disclaimer: This blog post is a simplified summary of published research for educational purposes. The accompanying illustration is artistic and does not depict actual model architectures, data, or experimental results. Always refer to the original paper for technical details.