Creating Meaningful Analogies for Real-World Problems with the TCSA Method + ChatGPT
In this article, I’ll present a structured method for creating meaningful analogies to real-world problems: TCSA (target-criteria-source-analogy). First, I’ll provide a short background on what I mean by “meaningful analogies”, the need to teach students analogical reasoning, and the difficulties with doing so. I’ll then explain how this method can be used to generate meaningful analogies to real-world problems. Finally, I’ll show how chatGPT (free version) can be used to make this process quicker as well as more effective.
My rules for using chatGPT to do anything that isn’t completely basic and simple, including to augment creative problem-solving, are as follows: (a) I already know how to do it well without using chatGPT, (b) I use structured prompts based on my knowledge and expertise, and (c) I use critical thinking to evaluate chatGPT’s responses. More specifically when it comes to creative problem-solving methods intended to generate novel and useful ideas/solution/products, I prefer to have the human intelligence (HI) generate responses (analogies, ideas, etc.) before prompting the AI to generate responses that may help uncover something the HI had missed.
Importantly, this method can be used effectively without chatGPT or any generative AI. In fact, the original method, which was very similiar to the one presented here, was developed and underwent initial evaluation before chatGPT was even released to the public, and had nothing to with generative AI. For more about my take on using chatGPT for creative ideation, you can read my exploration of this topic here.
Using and Teaching Analogical Reasoning for Problem-Solving
Analogical reasoning is a useful way of solving problems, as it helps find similarities between different situations; you can study an an existing solution to a problem (source problem, SP) that is similar to the problem you’re trying to solve (target problem, TP), and use your knowledge of the source problem and of its solution to guide your approach for addressing the target problem. A meaningful analogy is one which highlights similarities between the two problems that are relevant to addressing the target problem.
People have been using analogies for a long time to make important scientific and engineering breakthroughs. Analogies are used in many different fields, from computer science to industrial design, and they can even help improve conceptual learning in education. It’s therefore not surprising that teaching students how to create and use analogies for addressing problems, and fostering students’ creative thinking, is something many instructors in are interested in.
However, there’s a lack of such methods which can be applied easily, effectively, and reliably, especially with undergraduate students and with limited time for training by both instructors and students. With this need in mind, My colleague Dr. Deniz Marti from Harvard SEAS and I developed a method for structured creation of meaningful analogies to real-world problems. I deployed this method with faculty in two in-person workshops at the ASEE 2022 and TEE 2023 engineering education conferences.
My paper about the ASEE 2022 workshop with co-authors Dr. Deniz Marti (UCS) and Prof. Ed Crawley (MIT) was published in the proceedings for EDUNINE 2023, an IEEE conference. The method was originally called PC-SEA (problem-criterion-source-extract-analogy); the new version of the method, TCSA (target-criteria-source-analogy), is an improved iteration of PC-SEA based on lessons learned from its initial evaluation.
The Method
Target-Criteria-Source-Analogy (TCSA) is applied as follows:
(1) Target: Formulate a statement for the target problem (TP). A problem statement should ideally be short (~100 words), concise, and structured based on a much longer document. This will ensure that everyone involved in addressing the problem can recall its gist and understand it in the same way.
To do this, I recommend using a structured method for stating real-world problems, such as the 5W Technique which involves answering five questions about the problem:
1. Who is the group most affected by the problem? Key stakeholders
2. What are the adverse effects of the problem on our group?
3. Where does the problem occur? Physical and/or virtual spaces and locations
4. When does the problem occur? Events and periods
5. Why does the problem occur? Causes, enablers, and triggers
(2) 𝐂𝐫𝐢𝐭𝐞ria: Identify key usefulness criteria (KUCs) for the target problem (TP). KUCs are used to evaluate and compare ideas for addressing the TP.
(3) 𝐒𝐨𝐮𝐫𝐜𝐞: Find a source problem (SP) that shares as many KUCs as you can with the TP, but that unlike the TP has already been solved to a satisfactory degree. You can use YouTube or Wikipedia, for example — these are treasure troves for analogous problems. Formulate statements for the most suitable SPs, like you did for the TP.
(4) 𝐀𝐧𝐚𝐥𝐨𝐠𝐲: Create an analogy between the TP and SP through the KUCs. Since the analogy connects the TP and the SP based on multiple KUCs relevant for both problems, it will be meaningful.
Applying the Method to a Real-World Problem
Here’s a detailed example based on a real-world problem which I had students tackle in a spring 2022 course, before I developed TCSA. You can read an article about the problem-solving methods my students used in that course for problem structuring, systems thinking, and creative ideation here.
All the responses below were provided by chatGPT; however, as I describe in detail later, I recommend having human intelligence (HI) provide responses first for each stage, and only use chatGPT to augment those responses.
(1) Target: I created the TP statement based on a case description of the problem, which was also structured based on the five Ws. When using chatGPT to create the TP statement, I used the following sequence of prompts. The reason I didn’t use a single prompt for this stage was only because of the maximum character limit of chatGPT’s free version.
I am going to input description of a real-life problem, titled ‘target problem’. This input will have five parts, answering the five Ws: who, what, where, when, and why. Each part answers a question, as follows: “1. Who is the group most affected by the problem? Key stakeholders; 2. What are the adverse effects of the problem on our group?; 3. Where does the problem occur? Physical and/or virtual spaces and locations Gers; 4. When does the problem occur? Events and periods; 5. Why does the problem occur? Causes, enablers, and triggers”. I will now provide you with the description for “Who”. I want you to provide a summary between 150–200 characters of this description: “[content of section 1 from case description]”
And so on for each of the remaining four Ws: What? Where? When? Why?
Once chatGPT provided me with a short summary for each of the five Ws, I provided it with the following prompt:
Provide a summary of the problem in 300–400 characters, based on the following answers to Who, What, Where, When, and Why: “Who? [chatGPT’s summary]. What? [chatGPT’s summary]. Where? [chatGPT’s summary]. When? [chatGPT’s summary]. Why? [chatGPT’s summary].”
The above paragraph is the TP statement. We’re now ready to continue to the next step of TCSA, ‘Criteria’.
(2) 𝐂𝐫𝐢𝐭eria: I inputted the following prompt into chatGPT:
I want you to suggest five key usefulness criteria for evaluating solutions to the target problem. Provide a description for each criterion. ‘Usefulness’ in this context means the value or effectiveness of the solution.
We can select specific criteria from this list, add our own, or drill down into each criterion to create sub-criteria. Again, I suggest coming up with your own responses before asking chatGPT!
Now that we have out TP statement and our KUCs, we can turn to looking for analogous problems we can learn from.
(3) 𝐒𝐨𝐮𝐫𝐜𝐞: I prompted chatGPT as follows:
Describe five real-life problems that have been solved to a satisfactory degree and for which as many key usefulness criteria as possible belonging to the target problem are also relevant. These problems are titled ‘source problems’. Present each problem statement into a descriptive paragraph, like you did for the target problem statement, answering the five Ws for that source problem.
We’re now ready to create meaningful analogies which will help us ideate creatively on potential solutions to the TP.
(4) 𝐀𝐧𝐚𝐥𝐨𝐠𝐲: I prompted chatGPT as follows:
For each source problem you suggested, suggest one analogy with the target problem. Each analogy should connect the source problem with the target problem through as many of the joint key usefulness criteria as you can. Do not offer any solutions to the target problem, only provide an analogy between the source problem and the target problem.
With these meaningful analogies in mind, we can begin to think of creative ideas for addressing the problems — ideas which are both novel and useful.
Bonus: Creative Ideation with chatGPT
Just for fun, I asked chatGPT to suggest creative ideas based on the analogies it had provided:
Based on each analogy you provided, suggest creative ideas in engineering design for solving the target problem. Place emphasis on originality over usefulness, 2/3 weighting on originality and 1/3 on usefulness. For each idea, mention the relevant analogy it is based on.
While the first two ideas are worth considering, for the specific context of Ger District residents I’m also interested in ideas that emphasize affordability. So, I asked chatGPT for the following:
Based on each analogy you provided, suggest creative ideas in engineering design for solving the target problem. Place emphasis on affordability. For each idea, mention the relevant analogy it is based on.
Not bad! A few of these could be worth expounding on further. As always, I strongly suggested having the HI generate ideas first, and use chatGPT only to augment the process and provide summaries for the TP statement, potential SPs, and suggest analogies and ideas which may have been missed by people. I wouldn’t count on chatGPT for novel breakthroughs, but when the TP is well-structured like in the present example, chatGPT can be a lot more useful for creative problem-solving then when given a vague and/or badly researched problem.
Conclusion
Using TCSA, instructors can create all the required components for students to teach students creative problem-solving with analogies — target problem, key usefulness criteria, source problem/s, and analogies — and students can practice a structured way for creating analogies as part of creative ideation when addressing real-world problems. For more about the connection between the ‘structuredness’ of a problem and the usefulness of using chatGPT for ideation, read my Medium article about this topic.
Select References
Çubukcu, E. and Dündar, Ş. G., “Can creativity be taught? An empirical study on benefits of visual analogy in basic design education,” A| Z ITU Journal of the Faculty of Architecture, vol. 4, pp. 67–80, 2007.
D. W. Dahl and P. Moreau, “The influence and value of analogical thinking during new product ideation,” Journal of Marketing Research,vol. 39, pp. 47–60, 2002.
Jonassen, D. H. (2000). Toward a design theory of problem solving. Educational technology research and development, 48(4), 63–85.
D. Gentner, K. J. Holyoak, and B. N. Kokinov, “Introduction: The place of analogy in cognition,” The Analogical Mind: Perspectives from Cognitive Science, pp. 1–19, 2001.
N. Bonnardel and E. Marmèche, “Evocation processes by novice and expert designers: Towards stimulating analogical thinking,” Creativity and Innovation Management, vol. 13, pp. 176–186, 2004.
K. J. Holyoak and R. G. Morrison, “The Cambridge handbook of thinking and reasoning,” vol. 137, Cambridge: Cambridge University Press, 2005.
K. J. Holyoak, and P. Thagard, “Mental leaps: Analogy in creative thought,” MIT Press, 1996.
U. Goswami, “The development of reasoning by analogy,” In The Development of Thinking and Reasoning, pp. 61–82, Psychology Press, 2013.
R. Keller, C. M. Eckert, and P. J. Clarkson, “Using an engineering change methodology to support conceptual design,” Journal of Engineering Design, vol. 20, pp. 571–587, 2009.
K. Fu, J. Murphy, M. Yang, K. Otto, D. Jensen, and K. Wood, “Design-by-analogy: Experimental evaluation of a functional analogy search methodology for concept generation improvement,” Research in Engineering Design , vol. 26, pp. 77–95, 2015.
S. Keshwani and A.h Chakrabarti, “Towards automatic classification of description of analogies into SAPPhIRE constructs,” 6th International Conference on Research into Design, pp. 643–655, 2017.
Lavi, R., Marti, D., & Crawley, E. (2023, March). Creating analogies for design problem-solving: Initial evaluation of an engineering faculty workshop. In 2023 IEEE World Engineering Education Conference (EDUNINE) (pp. 01–05). IEEE.
J. S. Linsey, K. L. Wood, and A. B. Markman, “Modality and representation in analogy,” Ai Edam, vol. 22, pp. 85–100, 2008.
A. Schwering, K.-U. Kühnberger, and B. Kokinov, “Analogies — Integrating cognitive abilities”, Cognitive Systems Research, vol. 10, pp. 175–177, 2009.
