AI personalization does not require collecting everything. Motivation design can often work with smaller, purpose-bound, user-editable signals that reduce privacy risk while preserving relevance.
Engagement metrics show that users are doing something, not whether an AI product is helping them achieve a worthwhile outcome. Strong evals separate activity, progress, motivation quality, and harm.
AI companions can “level up” alongside users, but progression should expand useful capability and shared context—not simulate neediness, emotional rank, or an endless race toward deeper dependence.
Personalization fairness asks whether different users receive meaningfully different opportunities, prices, encouragement, or difficulty—and whether those differences are justified.
Contextual bandits can choose among motivational interventions while learning from feedback, but the reward function determines whether adaptation becomes helpful or manipulative.