The broken contract
Why AI coaches fail where human relationships succeed.
Developmental relationships are not transactional exchanges of information. They are reciprocal commitments. AI cannot enter that contract.
Listen
The broken contract
Every meaningful developmental relationship is built on a psychological contract: an implicit, reciprocal agreement between two parties about what each will invest and what each can expect. Denise Rousseau's foundational work describes this as promissory, implicit, and perceptual. It is not written down. It is felt. And it governs whether a person shows up to the relationship with genuine engagement or merely goes through the motions.
The effort heuristic is part of this, but it is not the whole story. Yes, people judge value by perceived effort. A hand-written letter lands differently than a generated one. But the deeper mechanism is about reciprocal obligation. When a coach rearranges their schedule for you, when a manager spends an hour preparing your review, when a mentor calls you out on something difficult, they are not just demonstrating effort. They are signalling commitment to your outcome. They are entering a social contract. And that contract creates a reciprocal obligation: if they are invested, you owe it to them, and to yourself, to invest in return.
This is what Gouldner described as the norm of reciprocity, and what social exchange theory positions as the foundation of all cooperative relationships. The investment is not just a signal of quality. It is the mechanism that generates the learner's own effort. The coach's commitment is what activates the learner's commitment. Remove the perceived investment, and you do not just reduce trust in the advice. You remove the social scaffolding that motivates the person to do the hard work of actually changing.
AI coaches break this contract at a structural level. The user knows, or senses, that the response was generated in milliseconds at near-zero marginal cost. The advice may be technically sound. The language may be empathetic. But there is no reciprocal commitment. No skin in the game. No one on the other side of the relationship who will notice if you disengage, who will feel the weight of your stagnation, who has staked their own professional reputation on your growth.
This makes AI-initiated interactions particularly hollow. Effective coaching often requires the coach to initiate: to notice something the learner has not noticed, to surface a pattern, to challenge an assumption. But when a notification from an AI coach arrives, there is no social weight behind it. It is indistinguishable from spam. No one chose to send it. No one will follow up if it is ignored.
The problem compounds from the learner's side. Data from platforms like Udemy consistently shows that only around 6% of users are genuine self-directed learners: people with a sufficiently strong metacognitive engine to self-diagnose, seek out what they need, and follow through without external structure. The other 94% require external prompting, accountability, and social motivation. Pure AI coaching asks exactly these people to do the thing they are least equipped to do: sustain their own developmental process with a partner they have no social reason to trust.
This is not a UX problem. Better prompting, a warmer tone, a more human-sounding voice: none of this repairs the underlying contract. The effort heuristic and the norm of reciprocity are not bugs in human cognition. They are features. They evolved to help us allocate attention and energy toward relationships that are genuinely invested in our outcomes. No amount of AI sophistication changes the fact that the investment is not there.
Sources
4 referencesRousseau, D.M.
1995
Psychological contracts as individual beliefs about reciprocal exchange agreements, promissory, implicit, and perceptual in nature.
Kruger et al.
2004
People use perceived effort as a heuristic for quality: the more effort invested, the more valuable the output is judged to be.
Gouldner, A.W.
1960
The norm of reciprocity as a universal social mechanism: when one party fulfils obligations, the other reciprocates with effort and commitment.
Jakesch et al.
2023
People perceive AI-generated content as less authentic, less trustworthy, and less worthy of engagement than human-produced content.