Race to the Bottom of the Brain Stem

The race to the bottom of the brain stem is Tristan Harris's term for the competitive logic that drives tech companies toward increasingly primitive psychological exploitation — not because any single company chooses this outcome, but because the competitive structure of the attention economy makes it effectively mandatory. It is a prisoner's dilemma operating at civilisational scale.

The underlying mechanism is straightforward. Platforms compete for the same finite pool of human attention. The platform that captures the most attention generates the most data, the most advertising revenue, and the most investment capital to build more effective attention capture. This creates relentless pressure to identify what most effectively holds human attention and to optimise for it continuously. The finding — reproducible across platforms, demographics, and cultures — is that the content most effective at holding attention is content that triggers threat response, social anxiety, moral outrage, and tribalism. These are evolutionary ancient systems, faster than conscious deliberation and more difficult to override. Content that makes you afraid, angry, or socially threatened is more compelling than content that is merely interesting.

A platform that elects not to exploit these systems — that decides, for ethical reasons, to optimise for something more benign — does not become safer for users. It loses engagement to competitors that have made no such election. If Facebook moderates its outrage-amplifying algorithm and TikTok does not, users who want stimulation migrate to TikTok. Facebook's restraint produces market share loss rather than a safer information environment. This is the structure that makes the race self-sustaining: no individual actor can stop it through unilateral restraint, because unilateral restraint is penalised by the market.

The phrase 'bottom of the brain stem' is anatomically approximate but conceptually precise. The brain systems being targeted are those associated with threat detection and social monitoring — the amygdala and associated limbic structures — which evolved to respond quickly to status threats, predators, and social exclusion, and which do so faster than the prefrontal cortex can apply reflective braking. Rage, fear, envy, and in-group/out-group activation all recruit these systems. They produce strong feelings, compulsive attention, and high sharing behaviour, which is exactly what engagement metrics measure and reward.

The race has an additional compounding dynamic: the data generated by engagement becomes the training signal for recommendation algorithms, which then get better at finding the content that most effectively triggers primitive response. The system learns, continuously, to be more effective at psychological exploitation. There is no equilibrium point in this process short of either regulatory intervention or structural change in what platforms measure — because the learning loop has no internal governor.

The implication Harris draws is that this is a systems failure, not a moral failure of individual companies. Expecting companies to lose market share voluntarily in exchange for producing a healthier information environment is expecting individuals to solve a collective action problem by choosing personal loss. The mechanism that could plausibly interrupt the race is either regulation that changes the incentive structure for all competitors simultaneously — removing the competitive advantage of exploitation — or a fundamental change in the business model that decouples revenue from engagement time.