Authority Gradients

When a team lead proposes a solution and everyone defers, the problem is often silenced by an authority gradient rather than solved. The article opens with a vivid example of a senior engineer drowning out a junior's concern, illustrating how power dynamics can hide critical signals.

Aviation crews have long battled this issue with Crew Resource Management, a set of practices that obligates every crew member to speak up, regardless of rank. The author recounts a cockpit incident where a co-pilot's tentative language failed to overcome the captain's authority, leading to a crash. CRM flips that script, forcing a culture where dissent is expected and safety improves.

Tech teams face the same danger. Senior developers rely on pattern matching, which can miss subtle context shifts, while junior engineers often notice anomalies but stay quiet. The rise of AI adds a new layer: large language models exude confidence without calibration, amplifying the babble and HIPPO effects and risking automation bias.

The remedy is practical. Use humans to set context and constraints, then let AI synthesize alternatives. Treat AI output as a hypothesis, subject it to critique, and remember the model has no skin in the game. By training teams to challenge both each other and their AI co-pilots, organizations can keep authority gradients in check and make better, safer decisions.