Week 27 Day 3: 'All You Gotta Do' Assumes Your Mental Model Is Shared

When you say 'all you gotta do is update the config file,' you are seeing the entire system in your mind: which config file, which parameters, what the downstream effects are, which services will need to be restarted, and what the rollback plan is if something breaks. The person you are talking to is seeing the phrase 'update the config file.' The gap between your mental model and their mental model is where the failure will happen.

Here is a diagnostic for mental model gaps. Take any task you have recently described as simple and write down everything you know about it. Not just the steps -- everything. The history of why it is set up this way. The last time someone tried to change it and what happened. The people who will need to be notified. The tests that should be run afterward. The failure modes you have seen before. Now look at what you actually communicated when you delegated the task. The ratio between what you know and what you said is the mental model gap. I did this exercise for a 'simple' database migration I delegated with the instruction 'all you gotta do is run the migration script.' Here is what I knew but did not say: the script has a known issue with tables that have foreign key constraints and requires a specific execution order. The staging environment does not match production, so the script must be tested against a production replica, not staging. The last migration caused a 20-minute service disruption because we did not disable the health check monitors, which triggered a cascade of false alerts. The DBA needs to be notified because the migration locks occur during execution and can block reads. Operations needs a 30-minute maintenance window approved. I knew all of this. I communicated none of it. The developer ran the migration script in staging, it worked fine, so they ran it in production during business hours without a maintenance window. The table locks blocked customer transactions for 12 minutes, the health monitors triggered alerts, the on-call engineer escalated, and I spent three hours cleaning up a mess that a two-minute context transfer would have prevented.

The mental model gap is formalized in cognitive science as the problem of 'shared mental models' (Cannon-Bowers, Salas, and Converse, 1993). Their research on team cognition demonstrates that team performance is directly predicted by the degree of overlap between individual members' mental models of the task, team processes, and equipment/technology. Critically, they found that mental model overlap does not develop naturally through proximity -- it requires explicit communication, shared training, or collaborative planning. The illusion that mental models are shared -- what Nickerson (1999) calls the 'false consensus effect in knowledge estimation' -- is one of the most robust findings in cognitive psychology. Nickerson found that people routinely overestimate the degree to which others share their knowledge, and that this overestimation increases with the person's level of expertise in the domain. The database migration example illustrates what Rasmussen (1983) calls the difference between 'skill-based,' 'rule-based,' and 'knowledge-based' performance. The expert leader operates at the skill-based level (intuitive, automatic responses to the migration task) while the novice developer operates at the knowledge-based level (conscious, effortful reasoning about each step). The 'all you gotta do' instruction attempts to transfer skill-based performance through a knowledge-based channel, which research on expertise (Chi, Glaser, and Farr, 1988) demonstrates is fundamentally inadequate because expert knowledge is organized in 'schemas' that compress information in ways that cannot be unpacked through simple verbal instruction.