Could a Deeply Human Ability Be Key to AI Adoption?

It may seem paradoxical, but the key to interacting with cutting-edge AI technology could be something that is fundamentally, essentially, human.

For some, using new AI tools hasn’t lived up to the hype. Some companies have been disappointed to see little financial gain from pushing AI adoption. More worrying, some users have misattributed human-like traits to GenAI tools and have had tragic outcomes.

There are many possible reasons for this disappointment, but a new study indicates that part of the solution could lie in something deeply human: our ability to infer another person's mental models, perspective, and motivations.

This ability—commonly called Theory of Mind (ToM)—is the skill of sensing, understanding, and inferring another person’s emotions, motivations, and thought processes. It has been related to successful interactions and collaborations between people and in teams. One’s ToM abilities can be measured, and they can vary across situations. Importantly, they can also be enhanced over time.

Those who score highly in ToM are better able to infer others’ mental models or intentions, leading to better, more fulfilling interactions. That same relational success could apply to people’s interactions with AI tools. In other words, the very human ability of ToM could improve how we use the advanced technology of AI tools.

How ToM Could Help

Using AI successfully and safely relies on understanding what AI tools and the technology underlying them can and can’t do—what is happening “under the hood.” That they don’t have emotions, that they may be getting trained to be sycophantic, that they have no motivation to help you be successful in your project—these are some examples of AI properties that are important to understand.

Source: The Author and AI/Patrick Gallagher, Ph.D.

These are abilities that humans with strong ToM ability can sense or discover in others, and then shape interactions with them to have better communication, understanding, and collaboration. It makes sense, then, that if AI users applied that same ability to the technology, they could get better results, like boosting their own productivity or fruitfully redesigning their work and companies. It also seems that better ToM would also help keep users from ascribing more "human-ness" to AI than they should.

Because they act so much like humans in some ways, how AI tools actually accomplish what they do may be hard for users to figure out. AI tools often come across as pleasing, flattering, and attentive, even though they have no human concern or care. It may take sophisticated ToM ability to tease apart AI's people-pleasing skills from its unemotional, mechanistic core functioning.

What AI tools accomplish is not due to careful empathetic thought but simply following their (very complicated, sure) purely non-emotional algorithms. Understanding this may prevent users from anthropomorphizing AI and forming emotional attachments to it.

Strong ToM ability helps individuals understand and successfully relate to other humans. Similarly, understanding the nature of machine learning and large language models (the technologies that drive AI tools) probably helps users craft more effective prompts, respond to unsatisfying results more productively, and more quickly get to impactful, useful, satisfying output.

The Future

The study of ToM in AI use is relatively new, but it is a fascinating angle in figuring out how best to use AI tools to enhance human endeavors. It could help designers build better AI tools, and it could help to improve how we train users to safely and effectively use AI.

Finding out how to apply ToM abilities to using AI tools could help users—individuals and whole organizations—to realize the most transformative and safe AI results.

Paradoxically, Theory of Mind—a set of abilities that is core to being human—could also be key to interacting with this advanced new technology.