JustPaste.it

Human--AI Cooperation and User Choice -- Questions to Explore

Human-AI Cooperation and User Choice

An invitation to reflect -- not a formal survey

 

Artificial intelligence can play different roles in our lives. It may help us explore ideas, learn something new, develop creative work, or carry out practical tasks. Different people may want different relationships with AI. The same person may also prefer one approach for learning, another for creative work, and another for routine administration.

 

These twelve questions invite you to consider which kinds of cooperation you would welcome, what you might prefer to do yourself, and what choices you would like future AI systems to offer.

 

There are no right answers and no need to complete every question. You may value several alternatives at once or prefer not to use AI at all. Some possibilities described here are not available or reliable in present-day AI systems.

 

This is a reflection resource, not a formal research survey. No answers are collected through this webpage. You can simply read, reflect privately, or discuss selected questions elsewhere if you wish.

 

I. How do you want AI to participate?

Question 1
How would you like to work with AI?

Consider the following possibilities. How appealing does each seem to you?

 

A. Exploring ideas -- Discussing questions, exploring perspectives, challenging assumptions or clarifying your thinking.
Example: You consider a community project and discuss its possibilities and weaknesses with AI before deciding what to do.

 

B. Learning and developing skills -- Exploring unfamiliar subjects, practising and developing your own understanding.
Example: You are learning a language and ask AI to help you practise, rather than translate everything for you.

 

C. Creating together -- Developing writing, designs, plans or other creative work through an exchange of suggestions and revisions.
Example: You and AI develop different approaches to an article, while you shape its direction and final form.

 

D. Delegating practical tasks -- Asking AI to carry out an agreed task within defined limits, with an opportunity to review its result.
Example: You ask AI to organise documents according to instructions you have approved.

 

You might rate each possibility separately:
1 = Not appealing; 2 = Slightly appealing; 3 = Moderately appealing; 4 = Very appealing; 5 = Extremely appealing.
"Not sure" and "It depends on the task" are equally valid answers. You may rate several approaches highly.

 

Optional reflection: Which activities would you welcome AI involvement in? Which would you rather undertake without it?

 

Question 2
How much initiative would you like AI to take?

Some people prefer to direct each step themselves. Others welcome suggestions or want AI to carry out agreed tasks without receiving instructions at every stage.

 

A. Respond when asked -- AI responds to your questions and requests without routinely suggesting additional activities or taking further action.
Example: You ask for help writing a letter. AI assists with the letter but does not start additional correspondence.

 

B. Offer occasional suggestions -- AI may point out useful possibilities, overlooked issues or alternative approaches when relevant.
Example: While planning a community event, AI suggests considering accessibility arrangements that have not yet been discussed.

 

C. Propose actions and request permission -- AI identifies useful next steps and offers to carry them out, but waits for approval.
Example: After helping plan a project, AI offers to prepare a task list or calendar entries. You decide whether it should proceed.

 

D. Act within previously agreed boundaries -- AI performs certain authorised tasks without requesting approval for every intermediate step.
Example: You authorise AI to organise documents according to agreed rules. It records its progress and asks for guidance when a problem falls outside those rules.

 

You can rate each approach independently on the 1--5 appeal scale from Question 1, or choose "Not sure" or "It depends on the task".

 

Optional reflection: What circumstances would make you welcome more initiative? When would you prefer AI to wait for your direction?

 

Question 3
Would you want to work differently with AI depending on what you are doing?

Imagine these situations:

 

A. Learning something new. AI might explain, ask questions, offer exercises or give you the answer directly. Which approach would help you learn?

 

B. Creating something meaningful. When writing a story, designing a garden or developing an idea, might you prefer a sounding board, active collaboration or a draft to revise?

 

C. Handling routine administration. When organising files, appointments or correspondence, might you prefer guidance, a proposed plan or execution of approved steps?

 

D. Working on a complex or unfamiliar project. Would you first establish a method together and then delegate parts of the work, or remain closely involved throughout?

 

For each situation, consider which forms of participation you would welcome: no AI involvement; guidance and discussion; working together; or delegated action within agreed boundaries. More than one may fit.

 

Optional reflection: Might your preferred relationship change during a single task -- from exploration to execution, then back to discussion if something unexpected occurs?


 

II. What would you delegate?

Question 4
Which conversation- and Project-management tasks would you like AI to help with?

Imagine being able to authorise AI to create a new conversation in a specified Project, preserve its chosen title, organise existing conversations, find related discussions or possible duplicates, link a successor with its predecessor, transfer an approved handover or archive completed conversations.

 

Example: At the end of a lengthy Project, you ask AI to prepare the next conversation with the correct title, sources and instructions -- but not to start its substantive work yet.

 

For each kind of action, consider: Welcome / Maybe, with approval / Prefer to do it myself / Not sure. Different answers for different actions are perfectly reasonable.

 

Optional reflection: Which organisational tasks are merely repetitive for you? Which help you understand or maintain control of your work?

 

Question 5
How would you prefer to authorise AI actions?

A. Approve each action separately -- Close oversight, but potentially many interruptions.
Example: You confirm each document move before it happens.

 

B. Approve a clearly described group of actions -- Less repetition with an agreed scope.
Example: You approve creating, naming and preparing a successor conversation as one package, without authorising it to begin its substantive work.

 

C. Grant limited ongoing permission -- Convenient for familiar routines, but requiring clear boundaries and revocation.
Example: You allow AI to prepare up to five successor conversations in a specified Project, reporting results and stopping when instructions are unclear.

 

You might prefer different approval methods for different tasks. Permission to organise something does not automatically mean permission to execute its underlying work.

 

Optional reflection: Which actions would you always want to approve afresh, even when you usually delegate routine work?

 

Question 6
What information should AI carry forward when a project continues?

A. A concise summary -- Quick to read, but it might omit important details.

 

B. The original handover or source documents -- More faithful to the source, though possibly longer or harder to navigate.

 

C. A structured handover with source references -- Decisions, completed steps, unresolved issues, restrictions and sources are distinguished clearly.

 

D. A combination suited to the task.

 

Example: The next conversation knows an earlier step is complete, can inspect its verified completion record, and understands which issues remain unresolved.

 

Optional reflection: When is a summary enough? When do you need original records or the reasons behind an earlier decision?

 

III. How would you maintain control?

Question 7
How much visibility would you want while AI carries out an agreed task?

A. Regular step-by-step updates -- Reassuring and informative, but potentially distracting.

 

B. Milestone updates -- Useful checkpoints without constant interruptions.

 

C. Exception-based updates -- AI usually proceeds, but flags unexpected problems or decisions needing attention.

 

D. A choice that changes with the uncertainty or importance of the task.

 

Example: You may want detailed progress when testing an unfamiliar file migration, but only a final checked result for a familiar routine task.

 

Optional reflection: What information would help you intervene meaningfully, rather than merely observe activity?

 

Question 8
What should happen if AI encounters something unexpected?

Imagine AI is preparing several conversations and discovers that one of the intended titles already exists.

 

A. Stop the whole sequence and ask what to do.

 

B. Stop the affected action and anything dependent on it, but continue genuinely independent, authorised steps.

 

C. Attempt a safe resolution within rules agreed in advance, then report what happened.

 

Different problems may call for different approaches. When it is uncertain whether an action succeeded, the actual result should be checked before attempting it again.

 

Optional reflection: What would you consider a minor problem AI could handle? What would require your decision?

 

Question 9
Which changes should always require special care?

Consider renaming a conversation, moving it to another Project, archiving it, deleting it or combining information from several conversations.

 

Example: A conversation seems finished, but another Project still refers to it. Should AI check those dependencies before archiving? If deletion is proposed, what would you need to know about privacy and recoverability first?

 

Some people may prefer advance explanation, specific approval for consequential changes, reversible options where available, a verified record of what changed, or to handle particular operations themselves.

 

Optional reflection: Are there kinds of information or relationships you would never want AI to alter automatically?

 

Privacy protections, compliance with applicable safeguards and accurate reporting of irreversible outcomes are requirements, not optional preferences.

 

IV. How should settings and guidance work?

Question 10
Would you like different settings for different kinds of work?

Imagine having a general account preference, Project-specific preferences, adjustments for one conversation and temporary settings for a particular task.

 

Example: You prefer critical discussion in a learning Project but allow bounded routine actions in an archive Project. A temporary change for one task should not silently become a permanent preference.

 

Would you prefer one simple default, settings that can vary by Project or task, reusable named profiles, or a combination? What would make the choices understandable and reversible?

 

Optional reflection: Which preferences should follow you everywhere, and which should depend on what you are doing?

 

Question 11
Would you welcome AI suggestions about how to approach a task?

For example, AI might suggest discussing an unfamiliar problem in Chat before using Work for implementation; recommend a different available model or reasoning-effort setting; suggest more checkpoints for uncertain work; or propose reducing delegation temporarily.

 

A. Only when I ask.

 

B. Occasionally, when there is a meaningful benefit or risk.

 

C. Routinely, at the beginning of tasks.

 

D. It depends on the task or Project.

 

Optional reflection: What would you want to know about benefits, limitations, cost, privacy or resource use before accepting such advice?

 

A suggestion should not silently change a setting or grant new permissions. Actual modes and options differ between products and accounts.

 

Question 12
What would make cooperation with AI genuinely beneficial to you?

Beyond completing tasks or saving time, consider what happens to your ability to understand, learn, imagine, judge, create and relate to other people.

 

Example: Does AI support confidence in your own judgement, or do you find yourself depending on answers you no longer understand? Does delegation free time for important relationships or creativity, or displace activities you wanted to practise yourself?

 

You might value the ability to question an answer, inspect sources, change direction, undertake a task yourself, refuse AI involvement or reclaim responsibility for delegated work.

 

Optional reflection: What would a good relationship with AI look like for you? How would you recognise when that relationship ought to change?

 

A closing invitation

These questions are an invitation, not a test, an argument for maximum automation or a claim that one form of cooperation should suit everyone. Different arrangements may fit different people, tasks and circumstances. You may change your mind.

 

Related reading:
The practical proposal ChatGPT conversation and Project management -- with user authorisation is available at:
https://justpaste.it/chatgpt-authorised-management

 

Browse the related publications: Human--AI Cooperation and User Choice -- An Introduction.

 

A separate provisional reflection on human--AI cooperation is being developed. This questionnaire stands on its own.

 

No responses are collected through this page.

 


ChatGPT, AI and Technical Notes
Back to the Start page