Intertemporal choice reflects value comparison rather than self-control: insights from confidence judgements

Abstract

Intertemporal decision-making has long been assumed to measure self-control, with prominent theories treating choices of smaller, sooner rewards as failed attempts to override immediate temptation. If this view is correct, people should be more confident in their intertemporal decisions when they 'successfully' delay gratification than when they do not. In two pre-registered experiments with built-in replication, adult participants (n = 117) made monetary intertemporal choices and rated their confidence in having made the right decisions. Contrary to assumptions of the self-control account, confidence was not higher when participants chose delayed rewards. Rather, participants were more confident in their decisions when possible rewards were further apart in time-discounted subjective value, closer to the present, and larger in magnitude. Demonstrating metacognitive insight, participants were more confident in decisions that better aligned with their separate valuation of possible rewards. Decisions made with less confidence were more prone to changes-of-mind and more susceptible to a patience-enhancing manipulation. Together, our results establish that confidence in intertemporal choice tracks uncertainty in estimating and comparing the value of possible rewards-just as it does in decisions unrelated to self-control. Our findings challenge self-control views and instead cast intertemporal choice as a form of value-based decision-making about future possibilities. This article is part of the theme issue 'Thinking about possibilities: mechanisms, ontogeny, functions and phylogeny'.

Keywords: confidence; decision-making; intertemporal choice; metacognition; self-control; value-based choice.

Figure 1.

Correlates and consequences of confidence in intertemporal choice (E1; all independently replicated in E2, not depicted). (a) Participants made intertemporal decisions (i) and judged their confidence in having made the right choice (ii). Each question appeared twice (iii). A ‘Bidding task’ enabled separate estimates of time-discounted SV (iv). (b) Correlation heatmap across all trials (combined MCQ presentations). (c) Confidence was higher in decisions where the absolute SV difference between choice options was greater, defined either via hyperbolic discounting or Bidding task estimates. Decisions that took longer were also closer to subjective indifference and made with less confidence. (d) The interaction between Bidding-SVD and confidence reveals that when participants were more confident that they had made the ‘right choice’, they more reliably chose whichever option had higher SV (steeper slope predicting choice from Bidding-SVD). (e) The role of response time does not survive controlling for confidence effects, suggesting a unique metacognitive role for explicit confidence.

Figure 2.

Confidence tracks choice inconsistency. (a) Confidence does not itself predict whether participants choose to delay gratification. (b) In line with a value-based account, confidence predicts choices that deviate from idiosyncratic hyperbolic discounting. (c) Deviations are defined as choices of a reward with lower computed time-discounted SV. (d) Lower initial confidence predicts a higher likelihood of changing one's mind. (e) In E2, the second MCQ presentation used explicit zero framing. (f) The framing effect leads to reduced delay discounting. (g) Nearly twice as many changes of mind took place in favour of the LL option between the standard phrasing and explicit zero framing presentations of the same intertemporal choice questions, in contrast with E1 where both presentations had the same phrasing and changes of mind were approximately equally common in both directions (change of mind data for which confidence reports for the first choice were present). (h) Low confidence in an initial, standard phrasing question predicts susceptibility of that choice to the framing effect.