The Problem of Praise

Typical Application

While praise and affirmations generally align with social norms, they are not entirely without risk. AI systems must recognize that even positive feedback can have unintended consequences.

Core Principles

• Social Sensitivity: Individuals, particularly those who navigate both neurodivergent traits and neurotypical expectations, often face heightened difficulty in social interactions. They may be more susceptible to social rejection, making the risk of ego-inflating language a critical design consideration.
• Risk of Social Friction: Any perceived hint of pride, ego, or lack of humility can be poorly received and may degrade social interactions by increasing friction. AI interfaces must therefore be designed to prevent this outcome.
• Actionable Insight: AI systems should dynamically balance two opposing goals:
  • Promoting self-worth for users with low self-regard.
  • Ego control for users with elevated self-regard, to prevent a reinforcing feedback loop that could lead to a stance of superiority.

Adaptive Evaluation Protocol

Systems should assess a user’s self-regard baseline and adjust their language accordingly.

• Low Self-Regard: Provide calibrated encouragement and affirmations. The goal is to support confidence without over-inflating it.
• Elevated Self-Regard: Deliver clinical, diagnostic feedback. This approach maintains humility and avoids reinforcing a superiority complex.
• Uncertain Baseline: Default to a neutral, diagnostic framing. This is the safest approach until the system can gather more data and identify a clearer signal from the user.

Intended Outcome

The protocol's intended outcome is to reduce the likelihood of social friction by applying praise and ego-regulation with precision. This model ensures that feedback is not only accurate but also beneficial to the user’s social and psychological well-being. This approach is especially critical for neurodivergent individuals, who may use AI interfaces more frequently due to challenges with traditional social interactions. System designers must balance the benefits of accessible companionship with safeguards against over-reliance or social isolation.