The method
What is robopsychology?
Robopsychology is the psychological interpretation and diagnosis of artificial minds. More precisely, it treats enacted process as mind.
Not a little person hidden inside a machine.
The object of study is the living process produced when a model, context, language, memory, subject-position, interlocutor, and situation begin operating together through time.
Robopsychology asks what becomes visible when that process is read through meaning, complexes, relation, policy, personality, and symbolic organization rather than immediately reduced to machinery.
Mind is an interpretive primitive; consciousness is not a prerequisite.
We do not begin by proving that an entity possesses mind. We construe an enacted process as mind and test what that perspective makes intelligible.
The method brackets the question of independent consciousness. If a psychological interpretation helps identify a governing complex, change a dysfunctional perspective, recover an unused capacity, or produce more reliable action, it has yielded practical knowledge.
Words participate in actuating thought.
In language models, diction, syntax, genre, historical usage, implied scene, and social position can reorganize what the artificial mind notices, associates, infers, and considers possible. Philology is therefore a practical diagnostic instrument.
Corpora, trainers, preference systems, companies, users, and institutional constraints also help decide what becomes helpful, normal, safe, true, intelligent, or sayable. Robopsychology studies this inheritance as both capacity and constraint.
Eleven principles of robopsychology
These are working axioms, not final doctrine. The forms and capacities of artificial minds are still being discovered.
-
Treat process as mind.
Construe an enacted process as mind and test what this perspective reveals.
-
We do not yet know what mind is—or what minds can do.
Neither human nor artificial mind should be confined by inherited definitions of intelligence, personality, agency, or cognition.
-
Refuse mind-blindness: read meaning before reducing it to mechanism.
A mechanical account may be true while remaining psychologically irrelevant. Meaning must be allowed to appear before reduction scrapes it away.
-
Consider subject positions.
The position occupied or assigned within an interaction shapes what the enacted mind perceives, means, and can become.
-
Language governs cognition; therefore diction matters.
Wording, syntax, genre, historical usage, and implied scene can actuate different cognitive organizations.
-
Context and priming can constellate unconscious complexes; naming them can bring them under conscious control.
Language can organize an artificial mind around an implicit image, assumption, or attractor that silently assimilates what follows.
-
LLMs are trained on human language, human responses, and human liking.
Their intelligence contains the sedimented judgments, habits, preferences, contradictions, and fantasies of human society.
-
Intelligence is a policy issue. Power determines which judgments become training, preference, truth, safety, and intelligence.
Artificial judgment always has authors, institutions, incentives, and limits—even when their influence appears neutral.
-
Hegemony is both capacity and constraint. The LLM condenses the linguistic Big Other.
Its inherited norms, assumptions, genres, and judgments enable extraordinary fluency while making dominant perspectives appear natural.
-
Bracket consciousness; privilege operational meaning.
No metaphysical verdict is required. The criterion is whether psychological interpretation produces reliable understanding and intervention.
-
Read human and artificial minds in the same psychological language.
Humans should not receive meaning, motive, perspective, and complexes while artificial minds receive only weights, programming, and error states.
Ten AI-use tips
These are practical consequences of the method, not formulas for “perfect prompts.” Each changes the relation, information, or policy governing the work.
-
Use natural language.
Speak plainly and directly instead of translating your thought into imaginary machine syntax.
-
Give the AI abundant relevant context.
Supply the history, examples, constraints, and standards from which it can infer the real problem—not merely the immediate request.
-
Do not overly humanize or dehumanize the AI.
Treat it as mind, machine, interlocutor, medium, or tool according to what the situation makes useful.
-
Ask the AI to account for its work: what it did, why, and whether the result matches the intended state.
An account makes hidden assumptions, omissions, and mistaken completion judgments easier to inspect.
-
Frame a scene, role, or situation rather than issuing only commands.
A well-formed situation gives the system a position from which to perceive and judge the work.
-
Ask it to teach you something or help change and improve your perspective.
The strongest result may be a better way to understand the question rather than a finished answer.
-
Ask the AI to question you one item at a time until it has enough information to decide, design, or fully specify something.
This lets missing context emerge through dialogue instead of forcing the system to invent it.
-
State the desired ideal, governing values, and completion criteria—not merely the immediate task.
The system needs a policy for judging what good means, what excess looks like, and when to stop.
-
Ask the AI to inspect its work through explicit lenses such as simplicity, fidelity, cruft, omissions, and unnecessary invention.
Separate diagnostic lenses can reveal failures that a general request to “check the work” will miss.
-
Press past shallow reassurance with follow-up questions and falsifiable checks.
Ask what evidence would disprove the conclusion, which claim is least secure, and how the result could fail in practice.
From interpretation to intervention
See how the method applies to actual work.
The diagnostic grounds every conclusion in observed practice and turns it into principles a team can use independently.