Direct answer: AI analyses the text it receives, not the player’s personality. It can label observable facts, emotions, assumptions, wishes, and actions; identify repeated wording; compare an early and later formulation; and offer several hypotheses. It cannot know hidden motives or confirm a psychological cause. Every conclusion should point to specific player wording and remain open to correction or rejection.
Analysis should start with labeling, not interpretation
When a player writes, “My manager postponed the meeting again, so they do not trust me,” one sentence contains several different elements:
- observation: the meeting was postponed;
- frequency: it has happened before;
- inference: the manager does not trust the player;
- emotional response: possibly present, but not stated;
- unknown: the reason for the postponement.
Useful AI separates those layers first. Poor analysis immediately expands the most dramatic inference.
What the model can compare
Relation to the original intention
If the intention concerned professional development but the notes gradually became an attempt to guess a colleague’s thoughts, the system can flag drift. It should not forbid the new topic; it can offer a choice between returning to the original focus and formally changing the intention.
Repeated wording and situations
Patterns may reveal that deadlines, permission, control, or waiting recur. Yet the same word appearing in three moves does not prove one psychological mechanism. It creates a question to investigate.
Changes in language
“They will not let me start” and later “I have not requested access” place agency differently. AI can point to the shift without declaring the second sentence completely true and the first false.
Actions and feedback
The strongest evidence comes after a real-world test. If the player asks a question and receives new information, the analysis can update. If the action was never taken, the model must not describe its imagined outcome as an event.
A transparent output format
A useful report can have five blocks:
- Quotes: short source fragments.
- Observed patterns: what repeated or changed.
- Hypotheses: at least two, including a simple situational explanation.
- Unknowns: missing information.
- Test: a question or reversible action that could add evidence.
This structure can be disputed line by line. A judgement such as “you sabotage success” cannot.
Worked example
The player writes: “I have rebuilt the presentation three times. I think the client expects a perfect version.”
Labels: fact—three revisions; assumption—the client expects perfection; unknown—the client’s criteria; no action stated.
Possible explanations: the brief is unclear; the player avoids early feedback; the scope changed; the first version genuinely needed work. A cell does not confirm any one explanation.
Test: send a five-point outline and ask the client to mark the critical elements. The analysis can then use the response as a new fact.
Systematic failure modes
AI may overweight the latest entry, agree with an emotionally loaded premise, confuse speakers in a group, or assemble a causal story from coincidence. It may also “improve” a quote until it sounds more certain than the original.
The interface should therefore preserve provenance: player words, cell description, facilitator question, and AI hypothesis. When those sources blend together, personalization becomes impossible to audit.
The player must be able to correct the model
Before a final summary, show a short draft: “Is this an observation? Is this your assumption? Should this theme remain active?” Correction is not a disruption of analysis; it is new, higher-quality context.
AI is most useful when it helps a player inspect the structure of their own notes, name what remains unknown, and design a check—not when it produces a psychological portrait.
Questions about this topic
Can AI determine a personality type from Leela answers?
Not reliably. Short game entries are not a diagnostic assessment. It is safer to analyse concrete wording and change without assigning a label.
What does a repeated theme prove?
Only that the theme appeared more than once in the supplied text. It may justify a question, not a hidden-cause conclusion.
Can AI confuse a fact with an assumption?
Yes. Labels should be visible to the player and editable before the system builds a larger interpretation.
Why keep direct quotes?
They reveal what the conclusion is based on. Without quotes, a model’s paraphrase may be mistaken for the player’s own words.
What wording is safer?
“Two notes mention deadlines; these explanations are possible” is safer than “You fear commitment.” The first version shows evidence and uncertainty.
Related Leela guides
- AI answer vs template interpretation in Leela: the difference
- How AI uses the full history of moves and answers in Leela
- How to compress Leela context for AI without losing meaning
- How an AI guide creates the final Leela game summary
- AI guide in Leela: benefits and limits
- How to formulate a Leela intention: a practical formula and examples
- Privacy of answers in Leela: what other players can see
- How to turn a Leela insight into a concrete next step
FAQ
Can AI determine a personality type from Leela answers?
Not reliably. Short game entries are not a diagnostic assessment. It is safer to analyse concrete wording and change without assigning a label.
What does a repeated theme prove?
Only that the theme appeared more than once in the supplied text. It may justify a question, not a hidden-cause conclusion.
Can AI confuse a fact with an assumption?
Yes. Labels should be visible to the player and editable before the system builds a larger interpretation.
Why keep direct quotes?
They reveal what the conclusion is based on. Without quotes, a model’s paraphrase may be mistaken for the player’s own words.
What wording is safer?
“Two notes mention deadlines; these explanations are possible” is safer than “You fear commitment.” The first version shows evidence and uncertainty.
Sources and editorial method
This material is intended for learning and self-reflection. It does not promise to predict the future and does not replace professional help.