I've cracked the secret of personal growth. Can I scale it with GPT-3?

TLDR: All personal growth can be boiled down to creating and resolving cognitive dissonance in an emotionally impactful manner. Seeking advice on whether AI/LLM can do it.

For years, I've been fascinated by the problem of promoting personal growth at scale, raising humanity's "sanity waterline".

An "AI coach" is ideally suited for this purpose.

The most straightforward approach is to train the model on recordings of coaching/therapy sessions + quality self-help materials.

However, a recent paper prompted me to believe there's a more nuanced approach based on how personal growth actually happens.
https://www.science.org/doi/10.1126/sciadv.abm0137

Outline:

1. Overview of the paper;
2. How does it relate to personal growth (also coaching and Buddhism);
3. The role of AI/LLM;
4. Open questions.

................................

The model in this paper has the power to explain personal growth computationally.

The idea of the paper is that we can conceptualize our beliefs as a network with some degree of dissonance. For example, the belief "I trust science" contradicts the belief "vaccines are a conspiracy", creating dissonance in the network.

It can also be characterized by "network interdependence" – the degree to which an individual is aware of the dissonance. Basically, it's one's self-awareness.

High dissonance + high awareness = strong incentive to resolve it.

Personal growth can be conceptualized as creating and resolving cognitive dissonance:

• Between conflicting beliefs = letting go of illusions;
• Between our beliefs and actions = building habits, behaviours and character strengths;
• Between our actions/beliefs and our needs = self-care, self-love, authentic goals, alignment with the world.

This idea resonates strongly with approaches suggested by modern coaching and Buddhism.

Coaching came to psychology from sports when it was observed that instead of teaching athletes the right moves, it's preferable to increase their awareness of their bodies so that they naturally improve their movements.

Increasing awareness = network interdependence is a key goal of coaching, achieved by asking questions and making observations.

Buddist philosophy is very similar: you cultivate a suffering-free mind by merely observing it and noticing how you create suffering for yourself.

Awareness -> higher perceived dissonance -> drive to resolve it -> change in beliefs and actions.

An LLM model can analyze one's cognitive model (by monitoring all conversations + asking questions) and change it in two ways:

• By pointing one's attention to the dissonance, questioning beliefs and actions;
• By suggesting more positive beliefs/actions which align with one's needs but conflict with negative beliefs/actions.

Though this sounds technical, it can be a very profound experience. Yesterday, I watched a 58-second video that pointed out that survivors of childhood trauma (like me) tend to overexplain themselves. I cried for several minutes – it felt like someone finally gave me permission to by myself without apologizing. AI (or a good therapist I couldn't afford before) could do it a decade ago.

I'm convinced such a solution can have a substantial positive impact. But is it technically possible in 2023?

1. Can the LLM build a model of someone's beliefs and behaviors?
2. Can it analyze dissonance within this model?
3. Can LLM suggest positive beliefs/actions that would fit specific individuals best on his network at the current point in their lives?
4. Can LLM do it in a compassionate manner that creates a transformative emotional experience – a requirement for lasting change?

Feedback would be highly appreciated! If you can't answer these questions, maybe you can point out where I can find these answers.

Thank you!