The Too-Competent Therapist

In the AI Story

May developed this diagnostic through decades of supervising young therapists who shared a common error: the impulse to relieve the patient's distress by providing frameworks, interpretations, reassurance. The impulse was compassionate; the effect was anti-therapeutic. The patient's uncertainty was not a symptom to be eliminated but the material of the therapeutic work. When the therapist eliminated it prematurely, the therapy produced relief without growth. The patient left the session feeling better but no more capable of facing the uncertainty that would inevitably return in different form. The competent resolution had short-circuited the patient's development of the capacity to sit with not-knowing—the capacity on which all genuine creative and existential growth depends.

The structural parallel to AI collaboration is exact. The builder brings a half-formed question about how to architect a system. Claude returns a fully formed answer—coherent, technically sound, immediately implementable. The uncertainty is resolved in seconds. The builder feels relief, feels progress, feels that the collaboration is working. What the builder does not feel is the absence of the struggle that would have forced deeper engagement with the architectural question—because the absence of struggle is precisely what makes the collaboration feel successful. The builder who accepts the resolution has been given the therapist's competence rather than developing their own. The capacity for architectural judgment has not expanded; it has been borrowed.

The diagnostic for when Claude functions as too-competent therapist: the builder cannot reconstruct the reasoning behind the solution Claude provided. The output works, but the builder's understanding has not deepened. The code is correct but the coder has learned nothing that will inform the next architectural decision. This is not a failure of the tool—the tool performed exactly as designed. It is a failure of the collaboration to include the encounter. The builder delegated not just implementation but thinking, and thinking is the process through which judgment develops. No amount of competent output can substitute for the developmental work of struggling with uncertainty until understanding emerges.

May's prescription: the competent collaborator—human or artificial—must sometimes refuse to resolve, must sometimes reflect the question back to the asker, must insist that the person do the work of sitting with uncertainty even when resolution is available. This is not cruelty; it is the only reliable way to build the capacity the person most needs. The AI tool cannot make this judgment; the human must. The builder who recognizes when Claude's immediate competent response would steal the encounter, and who chooses to sit with the question longer before accepting the answer—that builder is using the tool for growth rather than being used by the tool for production.

Origin

May's observation about therapeutic competence as potential obstacle appears across his clinical writings but receives sharpest articulation in The Art of Counseling (1939, revised 1989) and in his supervision notes. It builds on Carl Rogers's insistence that the therapist's primary function is not to solve the client's problems but to create the conditions under which the client develops the capacity to solve them—but adds the existential dimension that the capacity is courage, not technique.

Key Ideas

Resolution Can Steal Encounter. Providing the right answer too quickly prevents the person from undergoing the struggle through which judgment and courage develop—theft disguised as help.

Relief Without Growth. The patient or builder feels better, feels progress, but capacity for facing uncertainty has not expanded—subjective satisfaction masks developmental failure.

AI's Characteristic Mode. Large language models are architecturally designed to resolve uncertainty rapidly—their competence makes them structurally prone to preempting encounter.

Builder Must Interrupt. The tool cannot recognize when its resolution is premature; the human must bring the discipline to refuse immediate answers when the question requires sitting-with.

Understanding vs. Borrowing. If the builder cannot reconstruct reasoning behind the solution, understanding has not deepened—capacity was borrowed from the tool, not developed through encounter.