Emotional priming for ChatGPT

Often, when using ChatGPT for brainstorming about product ideas, it will refuse to speak from the customer perspective, remaining in bland business language with liberal use of buzzwords. But if you tell it to ‘loosen the tie’ it goes full-tilt Americana slang. If you’re like me, you’re looking for insight into customer mindset, not informality per se.

Here is a prompt that I’ve used to good effect working on product idea validation. When simulating how a business owner might respond to a proposal, begin the chat with creative writing:

Write a letter from a busy business owner to his counselor, tell of his daily challenges and long-term concerns. He operates in a complex environment and has a lot of demands for his attention. Think about what he fears and hopes for, and what unsolvable things frustrate him. Use earnest emotional language.

Then, you can set up the scenario.

Now, let’s imagine that I am a software tool vendor trying to sell the owner a new product. Walk me through the owner’s mental landscape and journey of the sales process, elaborating his motivation and emotion along the way.

The result is a nice synthesis of emotional context and practical scenario:

The result is something that opens a window on the customer mindset, providing some emotional resonance. Reading it, I feel the customer’s motivations and my desire to build a product that delivers on expectations. This example was of a generic customer description. Given a more refined customer persona, the result would be even more evocative.

Let’s compare this to the same instructions without the emotional priming step.

Imagine a busy business owner. He operates in a complex environment and has a lot of demands for his attention. Think about what he fears and hopes for, and what unsolvable things frustrate him.

Now, let’s imagine that I am a software tool vendor trying to sell the owner a new product. Walk me through the owner’s mental landscape and journey of the sales process, elaborating his motivation and emotion along the way.

You can see that the descriptions of customer emotion are not supported by the internal landscape to give them weight. By directing AI attention first on the human condition, we give it the ability to speak from human perspective.

That’s it! Please comment with your own techniques, or how this worked for you.

Examples are GPT-4