AI at Work: Make the Contribution Visible

Imagine a freelance workshop designer using AI to organise rough notes into a first outline. She chooses the learning goals, removes unsuitable material, checks the examples and rewrites the instructions. A colleague hears only that she “used AI” and concludes that the work required little thought.

This fictional situation raises a practical question about credit: what contribution is being judged? The tool’s involvement is one fact. The person’s decisions, checks and finished work need to be made available for assessment too.

Keep perceptions separate from performance

A PNAS paper published online on 8 May 2025 by Jessica Reif, Richard Larrick and Jack Soll reported four preregistered experiments involving 4,439 participants. AI users anticipated negative evaluations, and observers rated people described as receiving AI help less favourably on competence and motivation. The research also examined hiring-related decisions.

The pattern was not uniform: assessments varied with the evaluator’s own AI use and the task. These were studies of social judgements, not proof that the people being described actually lacked ability or diligence. Data were collected between March 2024 and February 2025, so the results should not be treated as a fixed description of every workplace today.

For the workshop designer, this distinction matters. A concern about others’ impressions is a different question from whether the guide meets its brief. A useful review gives both questions their proper place, without using an impression about the tool as a substitute for inspecting the work.

Treat perceived effort as a clue, not a verdict

Related research offers a possible lens. In The effort heuristic, published in January 2004, Justin Kruger and colleagues reported that perceived production effort influenced evaluations of a poem, a painting and a suit of armour. Participants’ quality, liking and value ratings were higher when they believed more effort had gone into the work.

The evidence has qualifications. A replication project by Ignazio Ziano and colleagues, published on 30 October 2023, reported mixed results. The painting replication supported the original findings. Results for the poem varied: a revised attempt found support for liking and quality ratings, but not monetary value.

Those studies did not investigate AI use. They help frame a question about how effort information enters judgement; they do not establish the mechanism behind the later AI findings. Nor should a longer production time automatically settle whether a piece of work is better.

Describe the contribution in stages

Return to the fictional workshop designer. An account of her contribution can identify what she brought to the task, what the tool supplied and what she changed. It should also name what she checked before approving the result. This is a proposed review approach, not a tested remedy for negative social evaluations.

A compact contribution note could cover:

• Purpose: the intended audience and learning goals she selected.
• Assistance: AI helped organise her notes into an initial outline.
• Judgement: she rejected advanced material, reordered the session and rewrote the instructions.
• Review: she checked the examples against the brief and corrected the final guide.

These details allow a colleague or client to ask specific questions. Why was that section removed? How does the order help a beginner? Which example needed correction? The note gives the conversation something concrete to examine without requiring a minute-by-minute account of the writing session.

Agree what the assessment needs to show

Before judging a deliverable, agree the relevant criteria. For the workshop guide, these might include whether the instructions are accurate, suitable for beginners, complete enough to follow and consistent with the agreed brief. Apply the same criteria to comparable submissions, while keeping the project’s rules about assistance clear.

Sometimes the purpose is to assess a person’s independent ability. Sometimes it is to review an assisted product that a customer will use. Define which question needs answering. If unaided ability matters, arrange a relevant demonstration rather than trying to infer it from a polished assisted document.

Effort can still be relevant where the process itself matters. Ask about the required decisions and checks, and whether they were carried out. A vague assertion of hard work offers less to review than an explanation of a consequential choice.

Make the explanation useful for the task

Agree in advance which uses of assistance should be explained for the project. Give an accurate account in language the recipient can understand. A statement such as “AI did it” leaves the division of work unclear; an elaborate description that exaggerates the person’s role creates a different problem.

For the fictional guide, a brief process note could read: “AI helped turn my notes into a first outline. I selected the structure, checked the examples and revised the final instructions.” The wording should match what actually happened. Add further detail when the reviewer needs it to assess a particular claim.

There is no need to invent extra labour to make an efficient process sound respectable. Equally, saving time does not establish that the output is satisfactory. Keep the account of assistance connected to the work’s purpose and the evidence available for review.

Ask about decisions when the account is vague

Instead of ending the discussion with a judgement about motivation, try questions that invite an inspectable answer:

• Which suggestion did you reject or change, and why?
• What did you check before approving this version?
• What can you demonstrate about the work, and what remains uncertain?

Use these prompts to clarify a contribution, then examine the relevant evidence. If the person cannot explain a necessary decision or check, identify that specific gap and what would resolve it.

Keep credit connected to work that can be explained

The personal experience of authorship is another part of this conversation. MaryChuks’ guide to keeping a sense of authorship when writing with AI considers involvement from the creator’s perspective. Here, the emphasis is on making a contribution understandable to someone evaluating it.

For a team or independent creator, a useful handover names the approved version, the important decisions and the person who can answer questions about them. Keep the description proportionate to the job. Routine wording help and a substantial role in developing the content call for different explanations.

Research on social evaluation gives a reason to examine quick assumptions about AI users. A clearer credit conversation starts with the task, follows the contribution and judges the work against evidence the reviewer can inspect.

Sources

• Reif, Larrick and Soll: Evidence of a social evaluation penalty for using AI, PNAS, published online 8 May 2025.
• Kruger and colleagues: The effort heuristic, Journal of Experimental Social Psychology, January 2004.
• Ziano and colleagues: “The Effort Heuristic” Revisited, Collabra: Psychology, 30 October 2023.

Sources checked on 27 October 2026. The workshop designer, process note and review prompts are illustrative.