How to Leverage Prompt Blocks To Foster Innovative Mindsets in Students

Making ChatGPT Your Design Thinking Co-Coach

Students doing on-site research for a business client … something AI can’t do.

This summer I'm touring schools in North Carolina with District C to coach students in design thinking by giving them an opportunity to pitch solutions to a problem for real business partners.

Yes, the process of developing innovative solutions is important … but we are really teaching them mindsets and tools for them to leverage when working in a group to solve complex problems.

The premise of District C and design thinking is that diversity drives innovation ... and this includes all ages ... even high school students. Given the right tools, students provide real value to our business partners.

That said, students can get stuck. Or sometimes think they've done "enough." I have found ChatGPT a useful tool for pointing out gaps and helping them realize what more they can do.

➡️ Check out one of my design thinking prompts here.

This week I explored using ChatGPT a bit more extensively, for example, helping students come up with more questions for their business interview and check-in, two important touch points that help them learn about the problem they are solving.

I simply swapped out the task prompt blocks.

[TASK] Develop a list of questions to research before the interviewing -business partner- to better understand the their problem.

I quickly realized, though, that ChatGPT can be quite convincing. I'm hypothesizing that when asking it to develop solutions, students are likely to clasp onto these, thinking that they are good.

Maybe they are, and maybe they are not. But really, they are just "hallucinations." ChatGPT is only guessing and knows very little about the problem or context. One could argue that students are also guessing and know very little about the problem, but we don't need two kinds of hallucinations.

The point of the design thinking process is to use empathy to learn about the problem from multiple perspectives.

Instead of fulling deploying ChatGPT into the design thinking process, I decided to take ChatGPT through the design thinking process.

Developing an Insight

Design thinking is certainly in GPT4's training set, but just asking for solutions doesn't necessarily implement the methodology.

Asking ChatGPT to take a solution and run it through the design thinking process will produce a convincing frame for the solution ... but it is all speculative. AI is not actually going through the design thinking process. It is taking a statistical guess on what that might look like.

So instead, I told ChatGPT what an insight is, and included student notes in the prompt.

[DEFINITION] An insight is a new way of thinking that helps us redefine the root problem so that we can build a solution for our partner.

[FORMAT] Use this sentence structure: Revlon thinks the problem is [blank], but really the problem is [blank].

I then cut and paste students' notes from the business interviews into the prompt, which delivered some interesting insights.

One of my co-coaches mentioned that ChatGPT can't deliver on the emotional and motivational aspects of insights ... so I asked it to revise the insights around emotion and motivation.

Here is what I got:

1. "Revlon sees the issue as high staff turnover, but the true issue lies in the absence of a nurturing culture, leaving employees feeling unvalued and overlooked."

2. "While Revlon identifies low employee engagement as the problem, the underlying issue is the lack of clear and transparent communication, causing employees to feel confused and disconnected."

3. "Revlon perceives the pressure on recruiting needs as the problem, but the real challenge is in refining the recruitment process to ensure candidates feel their skills are matched and valued in their roles."

4. "Revlon believes the inefficiencies in work to be the problem, but the root cause is the lack of competitive wages and rewarding incentives, leading to employees feeling undervalued and unmotivated."

5. "Revlon attributes the problem to financial difficulties, but the real issue is the need to foster a sense of job security among employees, which can boost their engagement and commitment."

I definitely can see this helping students come up with insights ... as long as they take high quality and detailed notes. Students will often have a good insight, but have difficulty putting it into words or fleshing it out.

The key is creating a prompt where students enter their notes and beginning ideas for insights, so that it is a collaboration with AI ... not a takeover.

Making AI a Co-Coach

My summer programs with District C are pretty intense ... just four days. This leaves little room to show them AI and give them the proper training ... especially students who are mostly new to using AI in meaningful ways.

(When asked if they have used AI yet, these students said, "Oh, you mean like in Snapchat?")

So instead of having them use AI, I created a "Robo Coach" document in their project folder. I structured two prompts ... one for each business partner, providing context for my coaching dialogues with the AI.

I also made sure to define specific District C terms, using what I call prompt blocks, or reusable and tagged paragraphs.

Then I gave students output in strategic places where their thinking process can be enhanced, mostly when coming up with ideas. For example, coming up with:

• More partner research angles.

• More questions for the client interviews.

• More ways to articulate an insight.

At one point, we were struggling with an unengaged student and even asked our AI co-coach to give us some suggestions on how to coach that student in positive ways. The results were useful!

Some good advice for working with disengaged students.

Once you have the structured data AI needs to be a coach, the output is much more likely to be useful for specific cases. This is why it is important to structure prompts and keep specific elements consistent across a team or an organization.

If everyone in District C is structuring their prompts differently (or not at all), the output is less likely to align with the organizations mission and values.

Going the Whole Way with AI

I continued to experiment with ChatGPT as a co-coach, seeing how far I could push its capabilities. It was clear that AI could be a powerful tool in the design thinking process, but only if used responsibly.

For fun, I went ahead and took ChatGPT through the entire process, asking it to choose which insight it thought would be most useful, taking it through an imaginary design thinking process, then creating a solution.

Then I asked AI to create a description for each slide, according to the most commonly used structure for pitch events.

• What we learned

• Insight

• How we came to insight (I add this one)

• Solution Prototype

• Action Plan

I then used those descriptions to draft an entire pitch deck. It wasn't bad ... and certainly plausible. But users must remember that it is simply guessing based on the data I provided.

➡️ See the full Gamma presentation here.

(If you would like to try Gamma with free credits, sign up with this link.)

But is this solution specifically tailored for our real business partner? Probably not.

This is what struck me after coaching this event. Our other partner told students that she expected some "out-of-the-box" solution, like having her hire a social media manager or posting more on social media.

She got the opposite ... a solution tailored for her. She didn't want to budget money for social media, or spend time posting herself. Both solutions took this into account

Yes, I (or students) can give ChatGPT these parameters, producing better output. But ultimately, it requires a human to make that connection. And users have to know to give such parameters to AI.

So what are we teaching?

The premise of District C is that we are teaching them AI-proof skills ... mindsets and tools that help them solve complex problems and collaborate in diverse teams.

Given the right frameworks, AI can replicate some of these mindsets and tools. Training a chatbot with these frameworks and structured data will make them even better.

So now I am thinking ... what is it we are teaching students? This has to be clear before we deploy ChatGPT in this context, or students will be more likely to hand over the process to AI.

So here is a list I compiled last week.

• Qualitative research on problem and partner

• Digging deeper into questions during interviews

• Making sure diverse voices are heard (not just yours or ChatGPT's)

• Taking notes and prioritizing information and data

• Developing an insight that connects directly to business contexts

• Constructing a usable prototype and action plan tailored for the partner

AI can certainly come up with good ideas or additional feedback when given the proper frameworks, but it can't do these for us.

My big takeway.

If there is is one thing students need to effectively use AI in future design thinking projects, it is this:

Students need to learn to make more structured notes. AI only works as good as the data we give it. If students aren't taking notes and don't know how to structure them, they won't be able to use AI in this context.

I noticed last week that students take terrible notes (or no notes at all).

What about you? What things do you think AI can or can't do in the design thinking process? What would you focus on teaching students? Leave a comment!