Collaboration Profile for ChatGPT
ChatGPT can already remember information about users and adapt its communication style. But there is another layer of personalization: how a user and ChatGPT prefer to work together.
I arrived at this idea through an extended experiment with ChatGPT personalization, using ChatGPT itself to analyze and refine the way we interact. The most useful result was not simply that ChatGPT learned more about me, but that a more effective way of working together gradually emerged: when to challenge my reasoning, when to suggest alternatives, when to go deeper in its analysis, how to communicate uncertainty, and how to use previous context.
While preparing this post, I discovered an existing proposal for a Persistent Collaboration Profile. Its author independently arrived at a similar conclusion after several years of using ChatGPT: it is valuable to preserve not only information about the user, but also the established way the user and ChatGPT work together.
I would like to propose an extension of that concept.
The existing proposal addresses the question:
How can we preserve an already established way of working between a user and ChatGPT?
I would add another:
How can we help users establish, configure, and gradually adapt that way of working in the first place?
Two Levels of Initial Calibration
Users may not always be able to determine in advance which way of interacting with ChatGPT will be most useful to them. A Collaboration Profile could therefore begin with one of two levels of initial calibration.
Quick Calibration
For users who prefer a simple starting point, this could be a short, optional, scenario-based questionnaire.
For example:
You propose a solution, but ChatGPT identifies a significant flaw in it. What should it do?
- Answer only the question you asked.
- Answer the question and briefly point out the issue.
- Explain the issue and suggest alternatives.
Users would not need to create complex instructions themselves. They could simply choose their preferred ChatGPT behavior in clear, practical scenarios.
Advanced Calibration
For users who want more control, an advanced setup could cover preferences such as depth of analysis, critical evaluation of reasoning, proactivity, handling of uncertainty, use of context, presentation of alternatives, and different rules for different types of tasks.
This mode could be particularly useful for professional use, but it should not be limited to professional users.
Both approaches would produce a single Collaboration Profile that the user could review and modify.
Adaptation Through Use
The initial settings would not necessarily need to remain static.
User preferences may change over time, and the most useful way of interacting with ChatGPT may depend on the task, context, and external circumstances. The Collaboration Profile should therefore be able to adapt to these changes.
For example, a user may initially choose concise answers but regularly request deeper analysis when making complex decisions.
ChatGPT could suggest:
“You prefer concise answers for simple questions, but when making decisions you usually ask for deeper analysis. Would you like me to automatically take the type of task into account?”
In this way, the Collaboration Profile could gradually be refined based on real-world use.
Significant changes should not happen silently. ChatGPT suggests an adaptation; the user accepts or rejects it.
How a Collaboration Profile Differs From Existing Personalization
The distinction can be expressed simply:
Memory: What does ChatGPT know about me?
Personality: How does ChatGPT communicate with me?
Collaboration Profile: How does ChatGPT work with me?
Contextual Adaptation
A Collaboration Profile could also account for context and task type. For example:
simple factual question → concise and direct answer;
important decision → multiple options, the consequences of each, key unknowns, and factors that could change the conclusion;
learning a new topic → explanation from first principles, examples, and gradually increasing complexity;
work-related task → a more formal approach, precision, constraint checking, and minimal irrelevant digressions;
creative task → more initiative, alternatives, and freedom to explore;
brainstorming → free generation of ideas without premature criticism, followed by a separate evaluation stage;
critical review of an idea → actively look for weaknesses and alternative explanations rather than simply reinforcing the initial hypothesis.
A single user does not necessarily need one fixed interaction style. A Collaboration Profile could define not only the user’s preferences, but also the conditions under which different preferences become relevant.
The Full Cycle
Combining initial calibration, adaptation, and the idea of a Persistent Collaboration Profile creates a complete cycle:
Quick or Advanced Calibration → Collaboration Profile → Real-World Use → Suggested Adaptation → User Confirmation → Persistence Across Conversations and Model Versions
Users would not need to develop an effective way of working with ChatGPT exclusively through prolonged trial and adjustment, while the collaboration patterns they have already established would not be lost over time.
The system should remain optional and transparent. Users should be able to review and modify their profile, accept or reject suggested adaptations, return to Advanced Calibration, reset the profile, or disable the feature entirely.
This is not intended to create a hidden psychological profile. The goal is much narrower: to adapt how ChatGPT and the user work together according to the user’s own preferences.
Conclusion
I arrived independently at the idea of a Collaboration Profile and discovered an existing proposal for preserving such a profile while preparing this post.
Rather than proposing a competing concept, I would like to suggest an extension:
Not only preserve an established way of working between a user and ChatGPT, but also help the user create it, test it in practice, and gradually adapt it.
The existing proposal effectively asks:
“How can I avoid losing the way I already work with ChatGPT?”
I would add another question:
“Why should users have to develop such a way of working entirely through prolonged trial and adjustment?”
A Collaboration Profile would not need to give ChatGPT new capabilities. Instead, it could make its existing proactive capabilities more systematic and better aligned with an individual user’s preferences.
Simple initial calibration + adaptation through use + persistence of the result.