23 Cake-Making Analogy for Setting Generative AI Guidelines/Ethics

Inspiration

The inspiration for this analogy came from a conversation I had with my cousin who works in Human Resources. We were discussing what kind of skills fresh graduates might be missing because of AI, and I asked her, as an analogy, whether staff who work with her needed to be able to bake cakes from scratch, or would there be instances where using cake mixes was the norm? I then expanded this analogy and asked my question on Twitter, with an image representing four ways of getting cake: from scratch, from a box, from a bakery, and from a supermarket (Image 1). The responses of Twitter/X users have since inspired many workshops and presentations I’ve done. I’ve also used the cake making analogy with my students and received responses from them on the same question on Instagram.

Image 1: Cake as a Metaphor for AI

What is the Cake-Making Analogy?

This analogy likens the use of “Generative AI” as a shortcut in an assignment to the different ways we make or get cake. One way is making a cake from scratch, buying all the ingredients and making it yourself. Another way is to get a readymade cake mix (like Betty Crocker) and build the cake. Another way is to buy a good quality cake readymade from a bakery. The last way is to get it from the supermarket, a low-quality version like a Twinkie.

Start with Cake

The activity starts by showing the audience photos of different ways of making a cake. Before mentioning AI, ask them when they would bake a cake from scratch, versus use a cake mix, buy from a bakery, or buy a Twinkie. People will respond differently, of course, depending on their baking skills, financial ability, and other factors. Possible questions to ask:

• Which of these ways of getting cake will produce a higher quality cake? Unless someone is a superb baker, a cake from a bakery is likely better quality – so why is that not always the right thing to do?
• What do people do when they don’t have time?
• What do people do when they don’t have or can’t afford certain ingredients?
• What about allergies, what do people do when baking for others with allergies?
• What emotional value does baking from scratch have?
• What are some of the advantages of baking from scratch? Participants may mention customization/personalization, emotional value, depending on occasion and who it’s for; enjoyment of the process; and learning and improving your own baking skills. We can ask then, what if your goal is to open your own bakery shop?
• Do you ever watch baking shows on TV? Do you notice how Nailed It bakers focus on decoration? They usually use ready-made cake mixes a lot of the time and then customize slightly, rather than bake from scratch. What are the implications of this approach?
• When someone gifts you with cake, is it necessary that they let you know whether/how they made it or where they bought it from? (inspired by Katie Conrad on Twitter)
• If our goal was a tea party, and cake was just one component of many other factors, how important would it be that the cake be homemade? (inspired by James Clay on Twitter)
• When is the process more important? When is the product more important? When is efficiency more important? When is personalization more important? When does it feel OK to bring the same thing everyone else brings?

Move onto Generative AI

Let’s apply this analogy to the use of generative AI. Depending on the maturity of the audience, you can invite them to compare what/how each approach to getting cake applies to doing assignments or using AI. If they need prompting, you could offer something like: Baking from scratch means doing something entirely on your own without any external support (including not using any generative AI). Buying from a bakery would be like getting a professional to write your paper or do your work, or in the case of AI, working with AI closely to complete the entire assignment. Baking from a box would be like using AI to write an outline or template but editing and filling in the gaps yourself. And the Twinkie would be like taking the first thing that comes out of your first prompt to AI.

Questions:

• What is the difference in quality between what you can bake from scratch and buy from a bakery? When might you do one and not the other? Translated to AI: what are some things you think AI can do better than you? What are some things where it is crucial you do them completely on your own in order to learn? What kind of support do you need to be able to do them on your own so you don’t end up seeking unauthorized help?
• What do we lose when we don’t make things from scratch? (in cakes, in learning)
• How might the harshness of a teacher’s judgment affect your decision to bake from scratch versus seek one of the shortcuts (inspired by Sybil from Twitter).
• When is it OK to use a cake mix (get a little help with outlining/template from AI)? When is “decorating” your main responsibility but the base cake can come through a shortcut? Where in this course might that be ethically acceptable?
• When might you feel like a Twinkie is sufficient? What elements of this course feel like they’re asking for something routine/vanilla, where everyone would create the same things and you’d feel tempted to submit a Twinkie? How can we change that?
• What would motivate you to do something yourself even if you had options to take shortcuts?
• What would indicate to you that the process of learning something is more important than getting the output “right”?
• What risks do you take when you bake/make from scratch? What risks do you take when you don’t bake/make from scratch?

Wrap-up by Creating Guidelines for Generative AI Going Forward

• Based on what we’ve discussed so far, what areas of this course are essential for students to “bake from scratch”?
• What factors may hinder students from baking from scratch even if required to (e.g., not having the skills, not having the ingredients handy, not having time) and what can we do to help them get there?
• When might it be OK to use a “ready-made cake mix” in this course (e.g. a template of some sort)? What is the equivalent of “ready-made cake mix” in this class?
• When might it be OK to buy a cake from a bakery? What is the equivalent of that for this class?
• When might it be OK to buy a supermarket cake like a Twinkie for this class? What is the equivalent of that for this class?
• When is it important to be transparent about how much you created on your own versus used AI or other sources?
• How can teachers inspire and motivate you to do work on your own?

Contexts in which AI Can Be Used

• Instructors with undergraduate students (and even high school and middle school students) to highlight to students what the core learning outcomes of their class are and help them reach, via dialogue, where “AI should be off limits” and areas where it should be OK to use AI, without hindering student learning.
• Educational developers with educators to help them highlight their core learning outcomes and where there may (or may not) be room to reasonably allow some use of generative AI without hindering student learning.
• In conversation with educators from different disciplines to highlight why different people will have different guidelines for generative AI use in their classes.

Reflections on Using the Cake-Making Analogy for AI

I’ve used this cake-making analogy in workshops to help faculty hone in on their core learning outcomes and identify particular assignments that they can either do away with completely, or allow some use of AI in. In all cases, faculty should make sure they focus their feedback/grades to students on the elements students do themselves. I’ve also used this analogy in my own class in order to discuss how the use of AI may support or hinder student learning. In both cases, it is important to bring up issues of inequities, such as the diversity of students’ baking capabilities (i.e., writing abilities and expertise in the subject) and variations of their financial abilities (i.e., resources to support them in doing the work).

When I used the analogy in class, I asked students to look up my original post on Instagram and respond there; then we discussed their responses in class. Only a couple of students commented on Instagram, and a few responded on our private class Slack. Some of the responses they gave were:

“I feel like it depends on the situation and context definitely. If i’m in a hurry and it’s a last minute decision it’ll depend on the proximity of my destination and how close the bakery or grocery shop would be so making the cake wouldn’t be an option unless its a special occasion where i’ve got time etc… but i feel unless its a known brand like twinkie etc… some people wouldn’t know the diff between using a ready-made mix or doing it from scratch which is kind of like AI? If u don’t use AI detectors. And for AI, maybe anything involving creativity and the students hobbies preferences would encourage them to do it themselves and make it more personal? Anything personal that wouldn’t seem like a chore would definitely encourage students i think!” – Isabella Matta

Isabella’s response highlights the importance of time, circumstance, and intrinsic motivation. She reminds us that if a student is doing something personal which they enjoy, they’ll want to do it “from scratch.” She also reminds us that in some cases, people might not “know the difference” between something done by AI versus a human, and for that kind of situation, people may be tempted to use AI, regardless of the guidelines.

Another student, Hassan, focuses on time and taste and mentions imperfections.

“I would make the cake at home from scratch if I have the time and I have something in mind to create and perfect it. On the other hand, I’d buy a fresh baked one if I’m in a rush and don’t have a lot of time and also I wouldn’t mind any imperfections that may arise in the taste because I didn’t bake it!” – Hassan Aboelela

Interestingly, he assumes imperfections are more likely if he didn’t bake it himself. He must think himself a really good baker! In class, we discussed what the response would be if he were someone who was NOT a very competent baker.

On Instagram, I asked him a follow up question:

Me: but you would never do it from a box or buy a ready-made preserved one (not fresh baked)?

Hassan: no I would do it from a box or a ready-made preserved one but it really depends on the situation. I will have a hierarchy based on two aspects: the time and the flavor. If I have a lot of time and care about the flavor, then I’ll do it from scratch. The more rushed I am, the more I’ll be inclined to buy it from outside. The hierarchy would be (from scratch, from a box, fresh from bakery, buy it from a grocery).

In Spring 2023, my students concluded that they should be allowed to use AI without disclosing in the following two ways:

1. Using AI to summarize readings to help them understand them more easily and read them faster, as long as they do not copy/paste the text into their written submissions. They argued that in this case they are not “submitting” an output that is AI generated, and therefore they do not need to account for using it. This would be like looking up a recipe for a cake, but then making up their own.

2. Using AI to do “scut work,” such as formatting references, as long as they brought the references themselves, input all the correct data, and checked the output. They argued this would be like spell check or grammar check, and not relevant to courses that aren’t particularly teaching language, writing, or citation skills.

We also agreed that any AI used verbatim needs to be cited. However, at that point we did not have many guidelines as to how to do that. Later, MLA’s guidelines came out, where AI would be cited along with the prompt used to produce the text.

We also agreed we would discuss timelines for assignments to ensure they were reasonable so students would not feel pressured for time.

Finally, we agreed that AI-produced content was never 100% trustworthy and would never be sufficient on its own. Therefore, wherever there was factual information, students needed to verify using credible sources and citations.

References

Aboelela, H. [@hassan_aboelela]. (2023, April 24). I would make the cake at home from scratch if I have the time and I have something in mind [Instagram post reply to Bali (2023b)]. Instagram. https://www.instagram.com/p/Crbbs4Fsyln/?utm_source=ig_web_copy_link

Bali, M. [@bali_maha]. (2023a, April 24). If we used cake as a metaphor for AI, I’m asking educators is: 1. When/where would it be acceptable for students [Tweet; image attached]. X. https://twitter.com/Bali_Maha/status/1650582997614092288?t=r3EfUYcldv58890x7wqiNg

Bali, M. [@edp05mab]. (2023b, April 24). If cake was a metaphor for #AI, when would you make it at home from scratch, versus make it at [Photograph]. Instagram.  https://www.instagram.com/p/Crbbs4Fsyln/?utm_source=ig_web_copy_link

Conrad, K. [@KatieConradKS]. (2023, April 24). I’d add that there is a cultural expectation that people disclose whether they bought a cake, made it from a [Tweet]. X. https://twitter.com/KatieConradKS/status/1650611229268799488?t=J1DdC6p8JfWuWr1KPVLtgQ

Clay, J. [@jamesclay]. (2023, April 25). This post did get me thinking. I think the answer depends on the outcome you want. If the outcome was [Tweet]. X. https://twitter.com/jamesclay/status/1650967611092418560?t=Pv_qUPtmIG5t-8xNcpCHag

Matta, I. [@isabellaa.matta]. (2023, April 24). I feel like it depends on the situation and context definitely. If i’m in a hurry and it’s a last [Instagram post reply to Bali (2023b)]. Instagram. https://www.instagram.com/p/Crbbs4Fsyln/?utm_source=ig_web_copy_link

Priebe, S. [@ihaveabug]. (2023, April 25). If the goal is learning, then home made encourages them to learn the process; however, then no matter how the [Tweet]. X. https://twitter.com/ihaveabug/status/1650903792987127811?t=cHApcjJYByRGcDj0m4nfUA

About the author

name: Maha Bali

institution: American University

Maha Bali is a professor of practice at the Center for Learning and Teaching at the American University in Cairo. She holds a PhD in Education and MEd in eLearning both from the University of Sheffield in the UK, and a BSc in Computer Science from the American University in Cairo. She is co-facilitator of Equity Unbound, which organizes intercultural global professional learning opportunities for educators and students. She blogs at https://blog.mahabali.me.