by Mike Critelli, 

Information Alone Is Not Enough

One of the challenges that has been top of mind to us as we build this browser-based AI-driven product is how to make it as effective as possible as it provides information and advice to users.

The mistake public health officials, the media, employers, educators, and many other leaders have repeatedly made is to assume that their job ends when they convey information about how we should act to maximize health.

We know from our personal and social experiences that even when family, friends, and other social influences are urging us to do the right thing, we fail to do so.

Why We Keep Choosing Unhealthy Habits

There are many reasons why our behaviors diverge from what we should be doing, especially when it relates to food.

We all know that unhealthy food with many unhealthy additives and preservatives is described as “comfort food” by individuals of all ages.

I know some individuals in their 70s who have survived life-threatening events, like heart attacks or strokes.  Although these events should have caused them to re-examine and change self-destructive behaviors, they provoke acts of denial and defiance by individuals who do not want to acknowledge their mortality.  They become even more determined to indulge and gorge on unhealthy foods or beverages.

They also may be engaged in unhealthy eating or drinking as a rebellion against the conformity they perceive when they follow a healthy crowd and practice unhealthy habits.

Standardized advice not only does not work with people who have these hardcore, but often unarticulated, motivations; it may achieve the opposite effect.

Motivational Interviewing: A Proven Approach

A well-designed AI agent learns what persuades us and the words and tones that deliver the persuasive punch.

At Pitney Bowes, our clinicians who regularly interacted with employees became certified in the Mayo Clinic motivational interviewing program.

They learned skills on how to do problem-solving with patients in a way that was non-judgmental and more focused on problem-solving.

They got great results.

Your AI Partner as Coach or Confidant

But the way we should think about an AI agent is as if it is a person with deep domain knowledge about motivational interviewing techniques that is always at our side.

It also may take the place of a loved one or a taskmaster that we respect.

In fact, one thing that we will consider as we build the tools to create a customized AI agent is how we incorporate the wisdom of others that we particularly respect.

Large language models incorporate the best resources available in rendering advice to us, but what we will do is to identify the sources of wisdom that each of us ranks the highest in a hierarchy of wise people.

Most of the time, we can identify those we consider to be role models, but, sometimes, a process needs to be activated to help us figure out which sources of insight are most likely to be those to which we respond.

Smarter Than the Average Counselor?

It learns about what moves us based on a far wider range of interactions than a professionally trained motivational interviewer can.

In fact, we can incorporate into a set of customized large language models the core principles of motivational interviewing.

Limitations—and Advantages—of AI

When we read about the limitations of AI, the most frequent answer is that they lack the empathy or the nuance of a human intervention.

Initially, that is true. But they have several advantages over the average human professional counselor:

• They are available 24x7.
  

• They can gather information about what motivates us, based on our daily behaviors and interactions, that we may not think to share with a human counselor. In fact, sometimes, we do not know what to share or may not want to share a reason for engaging in a seemingly self-destructive behavior.
  

• We may treat the AI advice giver as a source that does not make us feel that there is a stigma attached to our behavior.
  

AI + Human = A Daily Thought Partner

The best solution, at least initially, may be one informed by human guidance, but managed daily for a person by the AI “thought partner” we have helped users create.

Why This Matters Now

Well over half the American population needs interventions, since they are at various stages along a continuum from healthy to unhealthy and, for the most part, moving toward the unhealthy side of the continuum.

We want to move them in the other direction, or, at least, to keep them where they are.