What does it mean that building team epistemic practices?

Teaching your team the individual epistemic practices from this curriculum — calibrated confidence, assumption surfacing, perspective taking, evidence evaluation — creates collective capability that exceeds the sum of individual skills.

Example: An ML engineering team at a recommendation company had strong individual contributors but poor collective epistemic habits. Engineers would make confident predictions about model performance ('This will definitely improve click-through rate') without calibrating their confidence. The team would accept the loudest argument rather than evaluating evidence systematically. Assumptions about user behavior were rarely surfaced or tested. The team lead, Anya, introduced three epistemic practices from this curriculum into the team's workflow. First, calibrated confidence: before every A/B test, each team member independently predicted the outcome and assigned a probability. After enough predictions, the team reviewed their calibration — and discovered that predictions stated with 90% confidence were correct only 55% of the time. The visibility was sobering and transformative. Second, assumption surfacing: before each major project, the team listed the assumptions their plan depended on and classified each as 'tested,' 'testable,' or 'untested.' Untested assumptions became explicit risks. Third, pre-mortem analysis: before launch, the team imagined the project failing and worked backward to identify the most likely causes. Over six months, the team's project success rate improved from roughly one-in-three to two-in-three — not because their technical skills improved but because their collective epistemic practices reduced the frequency of preventable mistakes.

Try this: Introduce one epistemic practice to your team this week. Choose the one most relevant to your team's current weakness: (1) If your team makes overconfident predictions, introduce calibrated confidence — have each member predict the outcome of the current sprint's riskiest item with a probability, then review accuracy after the sprint. (2) If your team is surprised by failures, introduce pre-mortems — before the next launch, spend fifteen minutes imagining it failed and listing the causes. (3) If your team debates without resolving, introduce evidence mapping — before the next decision, list the evidence for and against each option on a shared document, rated by reliability. After the practice session, debrief: What did this practice reveal that our normal process misses? Should we make it a regular habit?