AI Co-Creation: Give the Finished Work a Fresh Review

You have rewritten the introduction, chosen the colours and negotiated every heading with an AI assistant. The finished workbook finally feels like yours. When a reader says that the first exercise is confusing, you want to explain how carefully you built it. That response would tell the reader about the work’s history. It would not yet answer their question about the exercise.

There is room for both pride and inspection in creative work. A useful distinction is between what making something means to its creator and what the finished object enables someone else to do. Research on the IKEA effect offers one reason to take that distinction seriously, provided we keep the evidence within its limits.

What the original experiments found

Michael Norton, Daniel Mochon and Dan Ariely’s The IKEA effect: When labor leads to love appeared online on 9 September 2011 and in the July 2012 issue of the Journal of Consumer Psychology. Across four studies involving assembled boxes, origami and Lego constructions, the authors investigated increased valuation of self-made products.

The authors’ earlier Harvard Business School working paper describes builders valuing a standardised storage box more highly than participants given an assembled box. Successful completion mattered: the reported valuation effect dissipated when participants failed to finish or assembled and then dismantled their creations.

These experiments concerned particular physical tasks and valuations. They did not test AI-generated workbooks, establish how every creator responds to feedback or show that personal effort makes a product worse. Applying the distinction to an AI-assisted project is an editorial proposal, not a finding from an AI experiment.

Separate the achievement from the reader’s task

Imagine a fictional illustrator called Nneka preparing a short drawing workbook with an assistant. She supplies the exercises, revises suggested instructions and chooses the final examples. After several sessions, she has a complete booklet. Finishing it is an achievement she is entitled to value.

For this proposed project, the reader’s task is narrower: open the first exercise, identify the required materials and begin drawing without additional explanation. That requirement can be reviewed separately from Nneka’s satisfaction with having completed the booklet. A product can be meaningful to its maker while one instruction still needs changing.

Suppose the exercise says to repeat the previous shading method, but the earlier page introduced two methods. The reviewer asks which one to use. Nneka’s account of why she selected the exercise does not resolve that ambiguity. A small revision naming the intended method would address the actual observation.

None of this requires concluding that Nneka is biased, unusually defensive or insufficiently skilled. The example simply gives the review a concrete object. Discuss the wording, the reader’s action and the proposed change, rather than diagnosing the person who made it.

Give the finished work a review brief

Before sharing the fictional workbook, Nneka could write down what she wants another person to check. A broad request for opinions leaves many possible standards in play: visual appeal, originality, clarity, enjoyment or readiness for sale. Choose the questions that matter for the present decision.

• Can the reader identify the materials needed for the first exercise?
• Can they point to the example the instructions refer to?
• Which words require an explanation from the creator?
• Does the final page make the intended next step clear?

These are suggested review questions, not a validated psychological intervention. Their purpose is to make the conversation specific enough to act on. Record what the reviewer actually did or asked, then consider the change that would address it.

For example, a note that the reviewer asked whether to use a pencil or charcoal is more informative for this brief than a score of seven out of ten. It names an unresolved decision. Nneka can then check whether the page omitted the material, used an unfamiliar term or deliberately allowed either option.

Use the assistant to examine a defined question

An assistant could help prepare alternative versions of the unclear instruction or organise comments into a revision list. This is a proposed use, not evidence that a particular model can judge artistic quality or predict a customer’s response. Keep the original comment available when considering its suggested interpretation.

Nneka might supply the two shading methods and ask for an instruction that names one unambiguously. She could also ask the assistant to list assumptions it has made about the reader. She would still need to check the result against the actual workbook and decide whether the instruction suits her teaching intention.

Do not ask an assistant’s praise to settle a disagreement with the reviewer. For this project, the question is whether the person could begin the exercise from the page supplied. A favourable description of Nneka’s dedication supplies no additional observation of that reading task.

MaryChuks.com’s article on preserving authorship when writing with AI explores how to keep consequential creative choices in your hands. A separate review can serve that intention: choose the work’s direction yourself, then examine whether the delivered version communicates it.

Keep a valued version and revise the working copy

There is no need to erase the finished version that marked an important milestone. In the fictional example, Nneka could archive that edition and make changes in a separate copy. This is a practical record-keeping suggestion. It does not establish that archiving reduces attachment or produces better feedback decisions.

The revision note might say that the original exercise referred to a previous method, the reviewer could not identify which method, and the revised sentence names it. That gives Nneka a clear reason for the edit. She can compare the versions without treating the first one as worthless.

She may also decide to retain a disputed choice. If the reviewer prefers realistic shading but the workbook deliberately teaches stylised marks, that preference may fall outside the review brief. Explain the intention and check whether the intended audience can recognise it. Taking feedback seriously does not require accepting every requested change.

Let pride and evidence answer different questions

The question of what you are proud to have made belongs to your account of the work. The question of whether a reader can use it needs observations about that reader and that task. Neither account has to displace the other.

For an AI-assisted project, celebrate the choices and effort that brought it into existence. Then give the finished version a specific review: who should use it, what should they be able to do and which evidence would support that judgement? That keeps the next revision connected to the work’s purpose.

Sources

Primary research checked on 30 October 2026. Nneka, the workbook and the review process are hypothetical. The proposed practices have not been tested for this article.

• Norton, Mochon and Ariely: The IKEA effect: When labor leads to love — Journal of Consumer Psychology, July 2012; first available online 9 September 2011.
• Harvard Business School Working Paper 11-091 — earlier 2011 working-paper version of the same research, not an independent replication.