A freelance illustrator wants to get better at negotiating with clients. An AI assistant offers an example in which a designer delivers a small set of product images early and sends the remaining illustrations the following week. The arrangement works for both sides. The example is clear, but will the illustrator recognise the same opportunity when a client asks for something other than a faster delivery?
Try putting that example next to a second one. A workshop organiser needs a venue within budget. The venue owner can offer a lower rate on a weekday, while the organiser can fill seats on a Thursday just as easily as on a Saturday. The objects and people change. The relationship is similar: each side cares more about a different part of the agreement. Seeing that relationship is the skill you want to carry into the next conversation.
What comparing cases can teach
In a 2003 paper in the Journal of Educational Psychology, Dedre Gentner, Jeffrey Loewenstein and Leigh Thompson tested ways to help novices learn negotiation strategies. In one experiment, people prompted to compare two cases transferred what they learnt to a new problem more successfully than people who studied the same cases separately. Another experiment found a benefit from stronger comparison guidance in a live negotiation exercise.
A 2003 study of negotiation teams by Loewenstein, Thompson and Gentner likewise found that teams who compared training cases were more likely to use a key strategy in a new negotiation than teams who analysed the same cases separately. These studies tested specific tasks and comparison instructions. They did not test AI tutors, and they cannot show that any pair of generated examples will improve learning. The practical idea to try is giving the learner an explicit comparison to make.
Choose examples that differ where it is safe to differ
Start by deciding what relationship the learner should notice. In the two fictional cases above, there are at least two terms under discussion, and the parties value those terms differently. The catalogue client mainly needs enough illustrations for launch day; the illustrator mainly needs time to finish the remainder. The venue owner values filling a quiet weekday; the organiser values meeting the budget and can move the event. In both cases, changing one term makes another easier to agree.
If the assistant generates two examples about illustrators dealing with identical launch schedules, it may be hard to tell whether you learnt the principle or just remembered the setting. If it changes the structure as well, there may be no useful comparison to make. Ask it to vary the setting and vocabulary while keeping the relationship the same. Read the result closely: a venue owner who cannot alter the rate, for instance, no longer fits the proposed example.
Do the matching before you read the explanation
Give yourself a short task for each case: name the parties, identify what each one values most and explain which change makes agreement possible. Then write a sentence that fits both. ‘Be flexible’ is too broad to tell you what to look for. ‘Find two negotiable terms the parties value differently’ points to something you can check in a new conversation.
- Ask for two short cases first. Tell the assistant to leave the shared principle unnamed until you have tried to explain it.
- Map the roles in both cases. Compare what each party needs, what it can change and how the trade works. If a detail has no equivalent, ask whether the detail matters.
- Keep an example that fails. If neither the price nor the date can move, the weekday arrangement no longer works. That counterexample helps define the limits of the rule.
Use a third case to see what travelled
Imagine a podcast producer needs a transcript for tomorrow’s launch. The transcriber can deliver the first ten minutes today, but the producer needs a complete transcript at launch. Splitting the delivery looks like the illustrator’s solution, yet it does not meet the producer’s priority. A better comparison would ask whether a different term can move: perhaps the producer can supply a clean script, or the launch timing can change. If neither is possible, the rule has revealed a constraint rather than a deal.
Try explaining that third case without the first two in view. Then ask the assistant which parts of your explanation are supported by the example and which depend on assumptions you introduced. This is related to mixing practice problems so you have to choose a method. Here the test is whether you recognise a shared relationship across different settings and can tell when it no longer holds.
For someone preparing a tutorial, the exercise also gives you a way to judge the examples you publish. Ask a reader to describe what the two cases have in common without borrowing your exact words. If they can repeat the slogan but cannot explain why the podcast arrangement fails, revise the comparison before adding another page of explanation. The test is whether the reader can use the idea when the familiar details are gone.
One way to make that test fair is to change the setting again while keeping the vocabulary simple. If the learner gets stuck on an unfamiliar term, explain the term and let them try again. You want to find out whether they can spot the trade, not whether they already know how a particular industry works.
Sources
- Dedre Gentner, Jeffrey Loewenstein and Leigh Thompson, Learning and Transfer: A General Role for Analogical Encoding (2003).
- Jeffrey Loewenstein, Leigh Thompson and Dedre Gentner, Analogical Learning in Negotiation Teams: Comparing Cases Promotes Learning and Transfer (2003).