Yet our AI systems, even the most advanced, focus almost exclusively on logical, step-by-step reasoning.
This is absolutely false.
We design them to explain every decision, show their work and follow clear patterns of deduction.
We are trying to design them to be able to explain their decisions and follow clear patterns of deduction, but we are still largely failing. In practice they often arrive at an answer in a flash (whether correct or incorrect), and this was almost universal for earlier models without the more recent development of "chain of thought".
Even in "reasoning" models there is plenty of evidence that they often still do have an answer largely determined before starting any "chain of thought" tokens and then make up reasons for it, sometimes including lies.
You're right! Modern AI often produces answers through pattern recognition rather than explicit reasoning. I should have been clearer that I was criticising our expectations of AI explanation, not necessarily how they actually function.
This actually strengthens my core argument. We're forcing explicit reasoning onto systems that may naturally operate more like intuition. We've privileged shared, verifiable reasoning over individualistic intuitive knowledge. Perhaps we've built AI to reflect how we wish humans reasoned rather than how we actually do. The irony is demanding explanations from AI that we don't require from human experts.
Recently, while playing cards with my family, I noticed something interesting about how I learn. While my husband and daughter approached the game through calculation and strategy, I relied on something less tangible, a sense of knowing that preceded my conscious reasoning. I won not by counting cards, but by developing an intuitive feel for the moves.
This experience made me reflect on something deeper.
What if we're missing half of what makes human intelligence remarkable?
Human intuition is the ability to arrive at accurate conclusions without explicit reasoning. This isn't just a cognitive shortcut, it's a sophisticated form of primitive intelligence with concrete neurobiological foundations involving the insular cortex, basal ganglia, and other specialised neural networks working in concert.
Yet our AI systems, even the most advanced, focus almost exclusively on logical, step-by-step reasoning. We design them to explain every decision, show their work and follow clear patterns of deduction, and not for the elegantly efficient "flash of knowing" that characterises human expertise.
Modern AI might perform trillions of calculations to produce an answer that appears seamless to users, while human experts often arrive at comparable conclusions with far less cognitive effort by leveraging intuition. This logical precision in AI comes at an enormous cost.
The efficiency gap becomes even more apparent in dynamic environments. While traditional AI must recalculate from first principles with each new data point, human intuition allows us to continuously update our understanding with minimal overhead.
If we want to build truly intuitive AI, we need to design systems that mirror the distributed, specialised nature of human cognition. This means moving beyond monolithic models toward integrated cognitive architectures with components that:
The technical foundations already exist. Bayesian neural networks, memory-augmented systems, Monte Carlo dropout techniques for uncertainty modelling. What's missing is an orchestration framework that integrates these elements into a cohesive architecture.
For too long, we've approached AI development through a false dichotomy. We either advocate for complete interpretability or black-box models without transparency. The path forward lies not in choosing between these approaches but in integrating them, just as human cognition seamlessly blends intuition and analysis.
Imagine designing AI systems that can respond instantaneously to familiar patterns with minimal computational overhead, recognise when a situation requires deeper analysis, build expertise through experience, and understand their own limitations.
Such systems wouldn't just be more efficient, they would fundamentally change our relationship with artificial intelligence, becoming true cognitive partners that complement human intelligence rather than merely mimicking its analytical aspects.
The goal isn't perfect replication of human cognition. It's thoughtful partnership. Not just machines that think, but machines that know when to trust a feeling.
Update: A commenter correctly pointed out that modern AI systems (particularly LLMs) often don't actually operate through explicit step-by-step reasoning, but rather through pattern recognition that may be somewhat analogous to intuition. My critique is better understood as focusing on our expectations and design goals for AI explanation rather than their actual functioning. This observation actually strengthens the core thesis that we need to reconsider our relationship with different forms of intelligence.
P.S. Accepting human intuition itself as a real form of intelligence remains a cultural challenge in many scientific and technical communities. We've long privileged explicit reasoning over implicit knowing, so I don't underestimate how difficult it will be to have ideas like this accepted in AI development. And we must acknowledge that we're far less forgiving of AI making intuitive mistakes than we are of human experts. Any implementation would need to begin with low-risk use cases, building trust and refining capabilities before approaching critical domains like healthcare or criminal justice.