Contribution
It proposes that the choice-induced preference change effect, the rise in liking for a chosen item and fall for a rejected one after a choice between items first rated equal, is formed during the decision rather than after it. The traditional explanation is post-decisional dissonance reduction (Festinger), with dlPFC adjusting vmPFC or striatal value at re-evaluation (p. 1). Voigt’s alternative: “preferences evolve dynamically as decisions arise, potentially as a mechanism to prevent stalemate situations in underdetermined decision scenarios” (p. 2).
Key insight
When existing values cannot separate two options, the decision system “extracts new information in the moment of choice via fixation on a particular item in order to construct new preferences guiding upcoming choices” (p. 2). The tie is broken by gaze, and gaze writes value.
Assumptions
- A serial architecture of five steps (pp. 2–3, Fig. 1; the text’s list
numbers them “(1) … (2) conflict detection, (2) value updating, (3) …,
(4)”, a typo for five):
- value computation, vmPFC (with dlPFC, vStr);
- conflict detection, dACC, when the value differential is too small;
- value updating, by fixation, with precuneus and dlPFC;
- value-based decision, dlPFC, once the differential suffices;
- updated value representation, hippocampus, for remembered choices.
- Fixation duration is causal in value-based choice, taken from earlier work: “Fixation duration plays a causal role in value-guided choice (e.g., Shimojo et al., 2003; Glaholt et al., 2009)” (p. 2), with Krajbich et al. 2010’s evidence-accumulation account.
- Preference changes do not occur for easy choices, “but only for hard decisions” (p. 2, from Voigt et al. 2019).
Key results
The article reports no new data. Its empirical basis, as it states it:
- From Voigt et al. 2019 (p. 2): preference changes “were predicted from activity in left dorsolateral prefrontal cortex and precuneus while making hard decisions”; “the total and first fixation duration for of an item was predictive of its choice and, importantly, its subsequent change in value (when controlling for choice)”; changes became behaviourally relevant “only for choices that were remembered”, with hippocampus activity. Hard choices took longer and engaged dACC (p. 2).
- The attentional step (pp. 2–3): “it is reasonable to assume that initially the precuneus … allocates spatial attention to the salient stimulus (i.e., bottom up) and forwards this information to the dlPFC”, which “‘holds’ choice-relevant information in working memory”. So “initial bottom-up shifts in spatial attention are explained by in-decisional precuneus activity controlling the reconstruction of new value information, which is forwarded to the dlPFC” (p. 3).
- A reinterpretation of dlPFC (p. 3): its role at re-evaluation in earlier studies “might simply reflect the computation of an upcoming value-based choice, but not necessarily the update of value itself as it was previously proposed”.
- Conclusion (p. 4): “seemingly self-determined, subjective cognitive concepts, such as our preferences, might be emergent consequences from the particulars of the decision scenario itself and brain networks underlying value-based decisions”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Choice-induced preference change arises during hard choices, not only after them | moderate | Voigt et al. 2019 (fMRI, eye-tracking), cited; one study |
| C2 | Fixation duration on an item predicts its choice and its later change in value, controlling for choice | moderate | Voigt et al. 2019, cited; correlational |
| C3 | Precuneus allocates attention bottom-up to the salient item and dlPFC holds it in working memory, together implementing the update | weak | ”it is reasonable to assume” (p. 3), assembled from studies of spatial attention and working memory |
| C4 | Long-term preference change requires the choice to be remembered, via hippocampus | moderate | three behavioural/imaging studies cited (p. 3) |
| C5 | Fixation-driven updating is an adaptive mechanism to break stalemates | weak | proposed, “might reflect”, “potentially” (pp. 2, 4) |
Concepts
- Choice-induced preference change — the post-choice spreading of ratings between items initially rated equal (p. 1).
- Hard decision — a choice in which “existing preferences are not sufficient to differentiate” the options (p. 1).
- Endogenous preference formation — change “in the absence of any additional external information about the choice options” (p. 1).
Connections
- Izuma et al. 2010 and 2015, Mengarelli et al. 2015 (tDCS over left PFC diminishing the effect), and Harmon-Jones et al. 2015: the dissonance-based, post-decisional account it revises.
- Krajbich, Armel and Rangel 2010 and Shimojo et al. 2003: gaze in value-based choice, the “causal role” it imports.
- Voigt, Murawski, Speer and Bode 2019, J. Neurosci.: its own evidence. Not held by the record.
Bearing on the record
- Jennings 2026 (LIT-1963, NOTE-1508, THEORY-1341): the use is
broadly faithful on the mechanism and unfaithful on two points.
- Faithful: vmPFC and ACC judging whether the value difference suffices; dlPFC and precuneus raising attention on one item, “allowing for both a decision and an updating of values in the memory system” (Jennings p. 25) matches steps 2–5. “To prevent stalemate” is Voigt’s phrase, though Voigt’s subject is preference change (“preferences evolve … potentially as a mechanism to prevent stalemate”, p. 2), not attention.
- Bottom-up, not top-down. Voigt’s attentional step is the precuneus allocating spatial attention “to the salient stimulus (i.e., bottom up)”, with dlPFC holding the result in working memory (p. 3). Jennings uses dlPFC feedback as the mark of fluid interests and of attention proper, as against salience-driven selection by fixed interests (manuscript pp. 31–32). On Voigt’s model the tie is broken by the salient item’s capture of attention, which on Jennings’s taxonomy is a fixed interest at work. That undercuts the step from Voigt to THEORY-1341’s criterion.
- Model, not demonstration. Jennings presents it as “a model for how such conflict is resolved in the brain” (p. 25), which is fair, but THEORY-1341’s “updates the chosen option’s value” rests on correlations (fixation predicts value change), plus a causal role for gaze imported from other studies. Voigt does not show that attention updates value; it proposes it, with correlational support.
- Jennings’s question “what determines whether attention falls on one option or the other?” (p. 26) has an answer in Voigt: salience, bottom up. Voigt’s own conclusion is deflationary about the self: preferences “might be emergent consequences from the particulars of the decision scenario itself” (p. 4), emergence in the weak sense.
- CLAIM-540. NOTE-1508 read the mechanism as a causal update, a mixture that changes because of the choice. Voigt’s evidence that the value changes because of the fixation, not merely with it, is the “controlling for choice” analysis of 2019 and imported gaze-manipulation results; within this article it is not an intervention.
- CLAIM-533. If the stalemate is broken by which item captures gaze, the outcome is decided by a fluctuation in salience, which is the “uncontrolled fluctuations and outside inputs” the claim says amplification does not tell apart from the agent’s effort.
- No instruction for machine-learning practice.
Limitations
- An Opinion article with no new data; the model is “tentative”.
- Its main empirical support is one study by the author, not read here.
- The precuneus–dlPFC division of labour is assumed (“it is reasonable to assume”, p. 3) from work on spatial attention and working memory, not tested on hard choices.
- The article does not say how the first fixation is determined beyond salience, nor what happens when neither item is more salient.
Open questions
- Does manipulating gaze during choices between equally liked items shift later preference, and does disrupting precuneus or dlPFC (e.g. by TMS) abolish the in-decision update?
- Is the tie broken by salience or by a goal, i.e. bottom up or top down? That is the question on which its use in THEORY-1341 turns.