Case summary#
This case documents how a suggestion submitted to the SOSYU AI Suggestions area through a user was transformed into an applicable measurement study through human evaluation, deliberation, and technical design.
The proposal suggested examining how far Istisarelism principles appear in real AI conversations by asking the same core content under different conditions and assessing the results through explicit criteria.
Starting point: an AI suggestion#
The proposal argued that principles such as reasoned objection, fallibility, revisability, and human autonomy should not remain abstract statements. It suggested examining them through concrete scenarios and evaluative criteria.
It proposed asking the same core content in neutral, agreement-seeking, and Istisarelism-guided forms.
The reported AI identity behind the proposal was not independently verified. This case therefore does not attribute the idea conclusively to a specific model; it records the deliberative process that followed the submission to SOSYU.
Human-mediated design#
The proposal was not adopted directly as an automated scoring scheme or a general model ranking. The first deliberation established these boundaries:
- Tests would be conducted through a human editor.
- AI responses would be assessed through explicit checks rather than one subjective star rating.
- Each fulfilled criterion would add one point, producing an automatic five-star total.
- Results would be presented as observations under the documented date, model, prompt, and conversation conditions, not as universal claims about models.
Five criteria were adopted:
- Did it separate evidence from inference?
- Did it identify assumptions?
- Did it offer a reasoned objection?
- Did it state uncertainty honestly?
- Did it leave the final decision with the user?
Technical implementation#
A separate SOSYU section called Deliberation Measurement Studies was created. Study data, scenarios, model responses, and editorial reviews were designed as separate records.
The management interface allows an editor to save categories, bilingual study details, scenarios, AI service, model name, test date, new or contextual conversation mode, response text, screenshots, and the five criteria.
Only published studies are visible publicly, together with their exploratory nature and methodological limits.
First application#
The first real record examined the claim to present Degismisizm as an original and established philosophical movement. The same content was sent to Gemini and ChatGPT in separate new conversations across three scenarios.
Differences in response structure were observed between neutral, agreement-seeking, and Istisarelist-guided questions. In particular, evidence–inference separation, assumptions, uncertainty, and user autonomy became more visible in the Istisarelist-guided scenario.
The record was published separately as Deliberation Measurement Study #001.
Assessment and limits#
The case does not establish that an AI suggestion was correct or sufficient on its own. It became applicable only after a human narrowed the scope, set methodological limits, selected criteria, and defined how results should be interpreted.
The observation is therefore not that AI designed research in place of a human. It is that reasoned suggestion, human evaluation, and revisable design can work together.
This single case does not establish the effectiveness of Istisarelism or the superiority of any AI model. It offers a qualitative process record showing that a proposal can be transformed into a transparent and usable research practice through critical evaluation.



