The Science of Intuition: When to Trust Your Gut

Cognitive biases are systematic tendencies to be wrong in predictable ways, not random errors. Living without biases is impossible, and most daily decisions aren't based on strict logic. Heuristics are simple rules of thumb that help us make decisions efficiently, while biases represent their systematic breakdowns. The representativeness bias demonstrates this: people choose coin sequences that look most random, even though all sequences are equally probable. This occurs because our brains prefer answering 'what is it like?' rather than calculating actual probabilities.

Heuristics are mental shortcuts that enable rapid decision-making but often lead to systematic errors; the substitution heuristic replaces difficult questions with easier ones, while the availability heuristic judges frequency based on recall ease, and the affect heuristic lets emotions unduly influence judgments, all of which can be mitigated through deliberate critical thinking.

Heuristics are mental shortcuts that the brain uses to process information quickly, while biases are the predictable patterns of automatic thought and behavior that emerge from these heuristics. The brain evolved these mechanisms because it needed to make rapid decisions to survive. Heuristics follow specific principles of brain architecture, including the need for speed and constant environmental modeling. These mechanisms are not flaws but evolved adaptations that served survival purposes. The term 'irrational' is problematic because it implies these behaviors are wrong, when in fact they serve evolutionary functions. Biases are so fundamental that they have been demonstrated in non-human primates, showing that fear of loss and risk-taking are basic aspects of mammalian cognition.

Cognitive psychology identifies systematic biases that lead to irrationality: availability bias (overestimating likelihood of memorable events like plane crashes), representativeness bias (judging probability by how well something matches stereotypes), and gambler's fallacy (believing random processes compensate to appear fair). These heuristics evolved for survival but fail in modern probabilistic reasoning, explaining many everyday errors in judgment.

Heuristics are mental shortcuts that the brain uses to simplify complex decision-making processes. These shortcuts, while efficient, can create systematic errors known as cognitive biases. For example, when asked about happiness and number of dates, there is no correlation. However, when the order of questions is reversed, the correlation jumps to 0.66. This demonstrates how people often substitute difficult questions with easier ones, leading to biased judgments. The brain is programmed to use these shortcuts, which can be both beneficial and problematic depending on the context.

Inside the Mobley factory, coating machine 14—the one transplanted from the original Essex works—the one that should have been numbered 13, still runs in near total darkness. Workers still navigate by red bulb torches. Greg Summers has spoken about conducting knowledge capture and succession planning, deliberately ensuring that the irreplaceable tacit knowledge of veteran workers, many in their 50s with decades of service, passes to a deliberately young research and development team. The knowledge of how to mix, coat, and finish photographic emulsion cannot be written down completely. Some of it lives only in the hands that have done it for 30 years. If that chain of transmission breaks, no amount of investment can rebuild it.

Tacit knowledge is non-verbal, experiential knowledge that cannot be easily articulated or transmitted through explicit instruction alone; it encompasses three main types: weak tacit knowledge (where observation and imitation are more efficient than verbal explanation, such as learning dance routines), body knowledge (skills embedded in physical practice like riding a bicycle that cannot be learned through verbal description), and social tacit knowledge (understanding gained through participation in communities, like language comprehension which requires being part of a living language community rather than simply having access to a database of rules); this type of knowledge is acquired through hands-on experience or social interaction rather than verbal instruction, and explains why experts often struggle to teach beginners effectively.

Much of what humans do and how humans reason is not written down—this is tacit knowledge. Mathematicians have intuition captured in conversations, lectures, and advising students, but only success stories are published. The precious data includes when someone tries something and it doesn't quite work, but they know how to fix it. Training AI on this tacit knowledge could unlock notable progress, but it would rely on human experts writing down their reasoning with fidelity. This approach wouldn't yield immediate explosive results but could be promising for long-term progress. The video concludes that AGI is not imminent but neither is AI all hype—it could be a combination of approaches that solves for abstract reasoning, with progress coming from merging deep planning paradigms with discretized search paradigms.

Tasks that are not easily automated are those requiring tacit knowledge - knowledge that humans cannot easily explain and is based on experience. This includes professional expertise such as a dermatologist identifying cancer from skin spots. While AI can analyze vast amounts of data to find patterns and produce similar results, it lacks genuine expertise, experience, and intuition. AI performs these tasks through different methods than humans, but it can achieve comparable outcomes by processing and analyzing data in ways that humans cannot.

Much of how products and processes actually work exists as tacit knowledge that cannot be captured in articles, write-ups, or videos. This knowledge must be communicated organically through demonstration and real-time interaction. The best teachers haven't prepared everything in advance—much of the magic comes from spontaneous moments when thoughts gel in real-time. Pair programming accelerates learning by revealing tools and tricks that would never be shared otherwise. Products contain more than tools—they embody how to think about work, including separating concerns and avoiding excessive integrations.

Modern neuroscience employs multiple techniques to reveal how brains make decisions. Electrophysiology measures electrical action potentials, revealing neurons in decision-making areas encode intermediate computational states between perception and action. Two-photon calcium imaging visualizes specific neuron populations, including inhibitory neurons that suppress other neurons. Whole-brain imaging across the dorsal cortex demonstrates that even simple visual decisions engage many brain areas working together. Decisions involve not just stimulus-response loops but also internal variables like biases and expectations. Each decision occurs within a particular behavioral state—brains cycle between high-activity 'up states' and low-activity 'down states'—significantly affecting how sensory signals are processed and decisions are made.

Research by Antonio Damasio and others revealed that emotions are fundamental to decision-making, not merely rational processes. Patients with frontal lobe lesions exhibit irrational behavior due to inability to moderate emotions. The limbic system includes the nucleus accumbens (positive emotions), insula (negative emotions), amygdala (danger sentinel), and hippocampus (emotional memory). These structures connect through the dopaminergic system from the ventral tegmental area. Decision-making requires integration between rational dorsolateral prefrontal cortex and emotional-limbic systems mediated by the ventromedial prefrontal cortex, with the orbitofrontal cortex integrating value signals from both domains.

Decision-making is fundamentally a selection between behavioral alternatives with different consequences, and the brain implements this through reinforcement learning mechanisms where dopaminergic neurons encode reward prediction errors that drive learning and behavior optimization. However, the brain does not function as a simple computational machine but as a highly parallel, distributed system with circular causalities across multiple levels, meaning decisions emerge from the dynamic interaction between brain, body, and environment rather than from isolated neural computations. This embodied perspective explains why decisions often involve intuitive 'gut feelings' and why cognitive biases may reflect evolutionary adaptations rather than computational errors.

Decision-making involves the brain's ability to choose one action over others, requiring evaluation of what is being sacrificed. The prefrontal cortex, located at the front of the brain, is responsible for decision-making and executive function. The ventromedial prefrontal cortex handles self-control and evaluation of choices. The brain uses memory to inform decisions, comparing past experiences with current options. Working memory holds recent information that influences immediate choices, while deeper memories linked to intentions and goals shape sustained decisions. When decisions are connected to meaningful purposes, they are more likely to be sustained over time.

The brain does not have one decision-making center. Instead, it is a network of different brain regions that constantly cooperate and compete. The prefrontal cortex is responsible for planning and control. The limbic system handles emotions, impulses, and reactions. Older, more primitive brain structures provide fast, automatic responses. Each wants something different, and what we call a 'decision' is the result of this internal tug-of-war.