Why Smart People Make Bad Decisions

Where intelligence helps.

Intelligence improves decision quality in conditions where the limiting constraint is analytical capability: complex data interpretation, pattern recognition across multiple variables, identification of non-obvious causal relationships. If the decision problem is primarily one of analytical capacity — the organization does not have enough processing power to handle the decision's complexity — intelligence directly addresses the constraint.

Where intelligence hurts.

Intelligence degrades decision quality in conditions where the limiting constraint is not analytical capacity but perspective diversity. Intelligent people are better at constructing convincing arguments for their pre-existing positions. They are more efficient at selectively retrieving evidence that supports their view. They are more capable of explaining away disconfirming evidence with sophisticated rationalisations. Research in this area — sometimes called 'myside bias' or 'motivated reasoning' — shows that the positive correlation between intelligence and reasoning quality disappears, and can invert, when the reasoning is in service of a pre-determined conclusion.

The compounding effect in leadership teams.

The problem is amplified in leadership teams, where the most intelligent members are also the most senior and therefore the most influential. Their sophisticated arguments for their positions are harder to challenge — not because the challenges are weak but because the cognitive overhead of contradicting a well-articulated, confidently held position is high. The team members who hold the dissenting view quietly discount their own judgment rather than sustaining an argument against someone who is clearly capable of defending their position. The intelligence that was supposed to improve the decision ends up suppressing the perspective diversity that good decisions require.

The structural response.

The most reliable response to the intelligence-as-liability problem is the same as for cognitive bias more broadly: structural perspective forcing. If the dissenting view has a formal role in the process — the assigned devil's advocate, the pre-mortem requirement, the Contrarian agent — it does not have to compete with the gravitational pull of the most articulate position in the room. Structure gives the dissenting view equal standing regardless of who holds it. This is one of the most important properties of well-designed AI deliberation: the Contrarian's challenge is not weakened by the intelligence of the person whose position it is challenging.