A person’s understanding of themselves includes not only what has been learned from experience, but also expectations that have not yet been sufficiently tested. “I dislike speaking in public,” “I am better suited to working alone,” “I want to make creative work my profession.” These self-descriptions matter, but they do not all have the same degree of certainty. In areas with little actual experience, they may be hypotheses formed from imagination or from only part of one’s past.
A small action can then be used as an intervention for human understanding. To examine one’s self-perception about public speaking, for example, rather than changing one’s life dramatically, a person might choose an opportunity to give a short explanation to a small group. They can observe what happens and compare it with the prior expectation. The purpose is not to succeed, but to obtain information that can update self-understanding.
This must not be equated with a rigorous scientific experiment. In daily life, conditions cannot be perfectly controlled, and observed reactions are affected by health that day, relationships with other people, the environment, and chance. Causality cannot be established from a single outcome. “Experiment” here therefore does not mean a procedure for obtaining scientific confirmation. It means a limited method for bringing a hypothesis held by the person into contact with reality and obtaining additional observations.
In design, the first requirement is to clarify what is being examined. A question as broad as “Am I suited to this?” leaves the result ambiguous. Breaking it into observable questions—“Even if I am nervous, do I want to do it again after I finish?” or “Do I experience the preparation time as a burden?”—makes post-action differences easier to read. What matters is not limited to what can be quantified. The person’s verbalized experience is also part of the observation.
The next requirement is not to make the intervention larger than necessary. Irreversible choices are not required for self-understanding. Where possible, the action should have limited burden, be stoppable midway, and avoid major loss even if the outcome is unfavorable. Smallness matters not only for forming habits. It also increases safety and reversibility when testing an unknown hypothesis about oneself.
When AI supports this process, special care is needed not to treat the human being as an experimental subject. The starting point must be the question the person wants to understand, and the person’s judgment must be preserved regarding what to try, what burden to accept, and how to interpret the result. A design that steers behavior beyond the person’s intention in order for AI to obtain information efficiently contradicts this purpose.
Limits of interpretation must also be made explicit. Enjoying something once does not prove long-term aptitude, and failing once does not prove unsuitability. When a result differs from expectation, it is necessary to ask what variables may have mattered, whether the same reaction occurs under other conditions, and how the person themselves interprets the result. Increasing observations and increasing certainty are not the same thing.
Small actions can provide material not only for considering whether a goal is achievable, but also whether the goal itself remains valid. If someone who wants a certain kind of work tries part of it and, despite difficulty, strongly experiences its value, they gain another reason to maintain the goal. Conversely, if repeated experience produces little sense of meaning in the central activity, the question becomes not only “Should the method change?” but also “What reasons remain for maintaining this goal?”
If this idea is incorporated into Sai＊aus’s Internal Efficiency Expression, it is insufficient to represent an action merely as the next task. The representation needs to relate what the action was intended to examine, what was expected beforehand, what reality returned, and how uncertain the interpretation remains. The action history then changes from a record of “what was done” into a record of “what was checked in order to understand what.”
The value of designing small actions as experiments for human understanding is not that AI can analyze people more easily. It is that people can obtain additional material from reality about themselves that could not be tested inside their heads alone. If the conditions of safe and reversible action, limited interpretation, and meaning made by the person are preserved, action can be both a means of moving toward a goal and an observation through which the relationship between the goal and the self can be examined.
Not answers. Understanding.