The literature on choice architecture — the systematic study of how the environment of choice shapes choice behavior — has produced excellent empirical work over the past two decades. The normative side of the conversation, I think, has lagged. We know a great deal about how defaults, framing, ordering, and salience shape decisions. We know less about when those shaping interventions are legitimate, and on what basis.
This post is an attempt to bring some structure to the normative question. I do not pretend to resolve it.
The standard cases
The standard cases in the choice-architecture literature involve interventions that are taken to be obviously legitimate. The classic example is organ donation: opt-out defaults dramatically increase donation rates, and the opt-out structure is held to be permissible because the underlying choice — to donate or not — remains available, and the social value of donation is high.
These cases have been used to motivate a general permissive view: choice-architecture interventions are legitimate so long as they do not foreclose the underlying choice. The technical term for this is libertarian paternalism, due to Thaler and Sunstein.
I think libertarian paternalism is a more limited framework than it is often treated as being. It works well for cases where the social value of the steered choice is clear, the alternative is preserved, and the agent can be assumed to share the value the architect is steering toward. It works less well for cases where any of those conditions fail.
The harder cases
Consider three harder cases.
Case 1. A commercial actor designs a choice architecture that steers consumers toward purchases that benefit the actor but not the consumer. The underlying choice (to purchase or not) is preserved. The intervention is permissible by libertarian paternalism’s standards.
Case 2. A regulator designs a choice architecture that steers citizens toward a policy preference the regulator favors but a significant minority of citizens reject. The underlying choice is preserved. The intervention is permissible by libertarian paternalism’s standards.
Case 3. A platform designs a choice architecture that steers users toward continued engagement with the platform, when the user’s reflective preference is to disengage. The underlying choice is preserved. The intervention is permissible by libertarian paternalism’s standards.
All three cases meet libertarian paternalism’s technical conditions. All three are, I think, morally problematic in ways the framework does not capture.
What the framework misses
What the framework misses, in my view, is the asymmetry of resources between the architect and the chooser. The architect has, in all three cases, more information about the choice environment than the chooser, more incentive to shape the choice in a particular direction, and more capacity to design subtle interventions whose mechanism is not transparent to the chooser. The fact that the chooser can, in principle, choose otherwise does not by itself make the architecture legitimate.
What we need, I think, is a richer normative framework that takes the architect-chooser asymmetry seriously. The framework should distinguish between cases where the asymmetry is acceptable (the regulator steering toward organ donation in a context where donation is widely valued) and cases where it is not (the commercial actor steering toward self-serving purchases).
What such a framework might look like
I will close with a sketch.
A choice-architectural intervention is permissible to the extent that:
(i) the architect’s interests align with the chooser’s reflective preferences;
(ii) the chooser, if informed about the architecture, would not reasonably object to it;
(iii) the intervention is reversible, in the sense that the chooser can disable or counteract it without disproportionate cost.
All three conditions admit of degree. None is by itself sufficient. But together, I think they give us better resources for evaluating choice architectures than libertarian paternalism’s simpler permissive criterion.
Whether this sketch holds up under pressure is another question. I am open to refinements, particularly from those who have thought harder about the political dimensions of choice architecture than I have.