Shaping or delegating decision-making
Palminteri and Wyart tackle the fundamental question of how technology changes who makes decisions and how.
Their framework distinguishing behavioural analysis, nudging, and boosting offers product leaders a vocabulary for thinking about when systems should shape human choices versus when they should delegate decisions entirely to algorithms.
The chapter's focus on AI as a decision-making delegate addresses one of the core tensions in product direction: the boundary between human judgment and algorithmic automation.
This work connects behavioural economics to organizational design, examining how technological capabilities create new forms of agency and responsibility.
For teams building AI-assisted products, it provides a theoretical foundation for thinking about the ethics and mechanics of human-machine decision partnerships.
Central argument
Palminteri and Wyart argue that the introduction of AI into decision environments is not merely a tools question but a structural one: it forces a choice between three distinct modes of intervention — behavioural analysis (understanding how decisions are made), nudging (steering choices while preserving agency), and boosting (enhancing the decision-maker's own capabilities). Their central thesis is that delegating decisions to algorithms is not a neutral or purely technical act but a reallocation of agency with ethical and organizational consequences. The framework positions AI as a potential decision-making delegate whose adoption requires explicit reasoning about which human capacities are being replaced, augmented, or bypassed.
Critique
The tripartite framework of analysis, nudging, and boosting is conceptually tidy but may underestimate how these modes collapse in practice — a system designed to boost human judgment can simultaneously nudge through interface design, making clean categorical attribution difficult and potentially misleading for practitioners trying to audit their own products. The essay also appears to treat the human-machine boundary as relatively stable and negotiable, when in high-velocity product environments that boundary is often set by engineering constraints, vendor decisions, or competitive pressure rather than deliberate ethical reasoning — a gap between the normative framework and organizational reality that the authors do not fully address.
Why it matters for product
For a CPO deciding how much autonomy to grant algorithmic systems in areas like prioritization, personalization, or resource allocation, the distinction between nudging and delegating is operationally critical: it determines where accountability sits, how teams are structured around decisions, and what metrics are even meaningful to track. The framework also has direct implications for AI-assisted product discovery — whether a recommendation system is boosting a PM's judgment or replacing it changes how you staff research, interpret outputs, and assign ownership over outcomes when things go wrong.