Your AI Mental Model: Bug or Feature?
What is Your AI Mental Model Costing You?
In a sense, every technology wave begins with confusion and a fight between the present and the past.
When new technology emerges, it's natural for us to understand it based on what we're already familiar with. These 'mental models' can sometimes make it challenging to see the new opportunities it presents. Over time, as our perspective shifts, what was once unclear often becomes obvious.
For example, when digital photography was invented by Steven Sasson at Kodak, it was viewed primarily through the lens of how to support the existing photo printing business. Others, not encumbered with that consideration realized that the attributes of digital photos could become a key enabler of new value.
Kodak created a digital camera, invested in the technology, and even understood that photos would be shared online. Where they failed was in realizing that online photo sharing was the new business, not just a way to expand the printing business.
HBR Article: https://hbr.org/2016/07/kodaks-downfall-wasnt-about-technology
New technologies offer more than just new solutions to old problems; they create entirely new possibilities. If we can recognise and move past ingrained mental models that hold us back and learn to see things differently, we can unlock a new technology's full potential early on.
Those who examine and refresh their mental models first, move faster to leverage the new technology's advantages and opportunities.
Are AI Hallucinations Really a "Bug"?
We see this happening today with Generative AI. For example, it's common to view factual inaccuracies, or hallucinations, as a flaw in the system. This perspective makes sense if we approach the AI like a search engine or a database, where the main goal is to retrieve a specific, correct answer.
Yet, these are Large Language Models, not Large Language Knowledge Bases. That's not to say they can't support that goal; but there are limitations and a different mental model required. This "Knowledge Base" mental model is like Kodak seeing digital cameras primarily as a way to sell more printing (a sane response in their mental model). This way of thinking, traps users within the old rules, blinding them to risks and new possibilities.
In fact, it's a recognised danger of Generative AI how convincing it can be when presenting incorrect information. One that is compounded by a single interface for use cases of both factual enquiry and creative intent. I even ran into this problem when doing research on historical examples of technology transition for this article. A misconception that I already had, which my AI tool was parroting back to me, would have been easy for me to accept as fact, because it aligned with what I already believed.
Turn a "Bug" into a "Feature" with a New Mental Model
A fresh way to think about Generative AI is to see its varied outputs not as mistakes, but as a source of boundless creative inspiration. Since the AI's outputs are based on a model of the world, not the world itself, its ability to provide 'alternative realities' can, when used correctly, be a major advantage.
Consider for example, a planning process: you can easily use AI to generate 5 plans in parallel, analyse their differences, identify your preferences among the versions, and use the insights you discover to optimise the next iteration.
When viewed this way, the AI's variability of responses becomes a feature that helps you produce higher quality results and innovate faster. It's a process that improves the final outcome and the insights you achieve improve your own skills and effectiveness.
Navigating a Blurred Reality
New technologies add to our options rather than replacing them. Most of us still have printed photos in our homes - my kids have polaroid cameras! The key is knowing which approach is best for the situation and switching our mindset for the task.
For example, we can treat AI as a research assistant that needs fact-checking, or as a creative partner for brainstorming new ideas. Developing this flexibility allows us to use new tools to complement and improve our existing ways of working.
Ultimately, the real challenge is not just using the tools differently, but fostering AI ready mental models to complement, enhance and "yes" sometimes replace our existing ways of thinking and working.
Try it Out
Of course, it’s one thing to hear about a new way of working, it’s quite another to experience it. So here’s a way to get started.
- Pick a Process: Think about a process you follow today that requires you to create an artifact (e.g., a plan, report, proposal, email). Choose something that is in part a creative endeavour that defines a future possibility rather than just reporting on the past.
- Generate in Parallel: Craft a prompt for your artifact and run it in three parallel instances of your preferred AI assistant to get multiple unique drafts.
Analyse the Variations: In a new chat window, paste all three of your initial drafts into this chat and instruct the AI to act as an analyst. Ask it to compare and contrast the drafts you provide, highlighting the key differences to help you identify your preferences. - Iterate with Insight: Use the AI's analysis to identify preferences and your own judgment to either craft a new, more refined prompt that asks for more of the best elements you've discovered, or to blend the best elements together.
- Rinse and repeat this cycle as required.
You will likely produce a better result. More importantly, you will learn something about what you consider defines a quality outcome for that process; an insight you can carry forward with you.
Tip: Don’t burn time to craft the perfect prompt to do this: LinkedIn is flooded with prompt advice, but for this quick experiment, just make the AI your prompt writing assistant. Fire up a new chat, explain your objective, and have it craft a prompt for you. That way you can get on with the creative part. If you want this to be part of your regular workflow, you can automate it.
Go try it. And if you uncover a new insight, share it!