We found that curiosity about the outcome of a lottery was enhanced for lotteries that were freely chosen, relative to lotteries that were equally preferred but not freely chosen

Choice Boosts Curiosity. Patricia Romero Verdugo et al. Psychological Science, October 26, 2022. https://doi.org/10.1177/09567976221082637

Abstract: In our connected era, we spend significant time and effort satisfying our curiosity. Often, we choose which information we seek, but sometimes the selection is made for us. We hypothesized that humans exhibit enhanced curiosity in the context of choice. We designed a task in which healthy participants saw two lotteries on each trial. On some trials, participants chose which lottery to play. On other trials, the lottery was selected for them. Participants then indicated their curiosity about the outcome of the to-be-played lottery via self-report ratings (Experiment 1, N = 34) or willingness-to-wait decisions (Experiment 2, N = 34). We found that participants exhibited higher curiosity ratings and greater willingness to wait for the outcome of lotteries they had chosen than for lotteries that had been selected for them (controlling for initial preference). This demonstrates that choice boosts curiosity, which may have implications for boosting learning, memory, and motivation.

Discussion

In the current set of experiments, we assessed whether freely choosing which information to sample increased curiosity. We found that curiosity about the outcome of a lottery was enhanced for lotteries that were freely chosen, relative to lotteries that were equally preferred but not freely chosen. In the absence of choice, curiosity was higher for lotteries that were preferred than for those that were not preferred. Furthermore, we investigated the effects of expected value and outcome uncertainty of lotteries on curiosity. In line with previous findings (Charpentier et al., 2018; Kobayashi et al., 2019; van Lieshout et al., 2018), our results showed that curiosity increased as a function of both of these factors. Interestingly, the effect of choice was independent of expected value and outcome uncertainty: Choice enhanced curiosity irrespective of how valuable the outcome was expected to be and of how much information could be gained from seeing it.

Choice boosted curiosity when participants reported their curiosity levels both explicitly (Experiment 1) and implicitly, by deciding to “pay” for information with their time (i.e., willingness to wait; Experiment 2). The fact that participants not only self-reported increased curiosity but were willing to give up a valuable resource (i.e., time) to see the outcome makes it unlikely that the effects were due to any inferred experimental-demand characteristics.

Our finding that choice increased curiosity generalizes the phenomenon of choice-induced preference change, observed in studies on value-based choice (Brehm, 1956; Izuma et al., 2010; Sharot et al., 2009, 2010), to the context of information seeking: Choice might boost the value of information just as it boosts the value of chosen options. Future studies could employ pre- and postchoice neuroeconomic preference tools to establish whether choice indeed alters the subjective value assigned to information.

Regarding what mechanisms underlie this choice-related curiosity boost, one possibility is that choice enhanced curiosity via an increase in (subjective) expected value. On the basis of previous evidence that perceived control inflates the subjective value of options proportionately to their objective value (Wang & Delgado, 2019), we initially predicted that an effect of choice on curiosity mediated by an increase in subjective expected value would result in an interaction between choice and expected value, whereas we observed an additive effect. Nevertheless, it is possible that choice enhanced curiosity by increasing subjective expected value equally for all levels of objective expected value. We cannot evaluate this possibility with our current data, but it is an interesting question that could be addressed in future studies using a revealed-preference procedure to assess participants’ subjective valuations of the chosen and unchosen vases.

Another possibility is that this choice-induced increase in curiosity resulted from participants’ drive to improve the quality of their decisions (i.e., learning to make better choices). Humans might learn through lifelong exposure that information and self-evaluation are especially useful under circumstances of agency (e.g., choice has been suggested to boost prediction errors; Cockburn et al., 2014). Hence, even though in our task, choice and no choice did not differ regarding learning opportunity, choice might generally boost capacity (or willingness) to learn, having enhanced curiosity in choice versus no choice in our experiment.

Alternatively, it could be that, rather than choice enhancing curiosity, no choice reduced it. In the choice condition, participants’ preferred lottery was always selected, but it was selected only half of the time in the no-choice condition. This could have generated an aversive association between no-choice and unfavored options or outcomes, in comparison with which choice appeared to boost curiosity. In this case, we would expect the effect of choice on curiosity to have increased throughout the task, as participants learned the contingencies and developed the aversive association for the no-choice condition. Follow-up analyses suggest that the effect of choice on curiosity was stable throughout the task (see the Supplemental Material), rendering this explanation unlikely. It remains possible that participants exhibited an aversion toward the no-choice condition right from the start. Further investigation is needed to assess to what extent choice increases curiosity and no-choice decreases it, for instance in a between-subjects experiment, in which one group performs the current task (choice vs. no choice) and another a no-choice task (otherwise matched).

An intriguing question for future work pertains to the neural mechanisms underlying the effect of the choice-related curiosity boost. Given the link between midbrain dopamine neuronal firing and information-prediction errors (Bromberg-Martin & Hikosaka, 2011), it is plausible that the mechanism suggested to underlie choice-induced preference change, involving feedback projections from the striatum to the midbrain (Cockburn et al., 2014), could also account for choice-induced increases in curiosity.

In addition, the noradrenaline system might play a role, given its implication in modulating arousal, which is likely enhanced in conditions of increased autonomy (Howells et al., 2010; Sara & Bouret, 2012; Varazzani et al., 2015). A role for noradrenaline in information seeking is also supported by recent evidence suggesting pharmacological blockade of noradrenergic-receptor stimulation with propranolol, decreased information gathering in a task in which participants uncovered cards until they felt confident to guess a predominant feature (Hauser et al., 2018). Hence, the increased autonomy experienced under choice in our task may have further implicated the noradrenergic system, boosting participants’ information-seeking drive.

Preference enhanced curiosity. In other words, for lotteries selected without participants having a choice, they were more curious about lotteries that they preferred than those they did not, suggesting that preference itself boosts the perceived value of information. Alternatively, it is possible that the mechanism underlying the effect of preference on curiosity was akin to that of choice, given that in our experiments, participants had to provide a preference indication (i.e., engaging in a comparison process, making a decision, and executing a response), which they might have experienced as a kind of choice.

Curiosity increased as a function of expected value. In other words, the more positive participants expected the information to be, the more curious they were. This aligns with previous findings that humans prefer information about future desirable outcomes over information about undesirable outcomes (Kobayashi et al., 2019) and are willing to pay for the positive information (Charpentier et al., 2018). This effect of expected value was more marked than in previous studies with a similar paradigm but without the valuation and choice phases (van Lieshout, de Lange, & Cools, 2021; van Lieshout et al., 2018; van Lieshout, Traast, et al., 2021). Unlike in the previous studies, participants in the current study were instructed to make a value-based comparison between pairs of lotteries to indicate their preferred option and to make a selection. These task features likely increased the salience of the expected value of the lotteries. Other characteristics of the task might have boosted participants’ attention and/or engagement further, including the longer duration of the lottery presentation (up to 11 s, vs. 3 s in previous studies) as well as the fact that participants got to continue with their preferred lotteries in the majority of trials (75%), whereas in previous studies, participants did not have a choice between lotteries. Surprisingly, we found that in lotteries with higher expected value, the effect of preference on implicit curiosity diminished. One possible explanation is that this higher expected value became more salient than other features of the trial, driving participants to focus their attention on the points they could win and less so on whether their preferred lottery was selected. Nevertheless, it is not clear to us why this would be the case when curiosity was assessed implicitly but not explicitly.

Our findings are in line with recent evidence that autonomy boosts curiosity: Participants who watched a video of their choice (from a given set) self-reported higher interest in the topic than those who watched a video without a choice (Schutte & Malouff, 2019), and participants bid larger amounts for lotteries they chose than those they had not chosen (Jiwa et al., 2021). However, in these recent studies, autonomy was confounded with preference. In the earlier study (Schutte & Malouff, 2019), the researchers did not consider participants’ preferences when assigning the options. This resulted in participants in the choice condition likely choosing their preferred option, and participants in the no-choice condition likely receiving their not-preferred option (with a probability of .66 because there were three options). In the more recent study (Jiwa et al., 2021), participants always made a choice, but half of the times, their choice was vetoed (resulting in a loss of agency). In agency trials, participants always received their chosen (presumably preferred) option, whereas in no-agency trials, they always received an unchosen (presumably not preferred) option. These designs did not enable dissociation between preference and choice; hence, the supposed effect of choice on curiosity also included a (confounded) effect of preference. Our study goes beyond these previous efforts by quantifying the effects of choice and preference separately, by demonstrating that these two factors boost curiosity independently, and by reporting these effects using both explicit (ratings) and implicit (willingness-to-wait) measures of curiosity.

Our paradigm enabled us to manipulate choice, outcome uncertainty, and expected value in a quantitative and controlled fashion, allowing us to draw inferences independently of extraneous variables. Although other paradigms relate more closely to real-world situations (e.g., trivia paradigms; Kang et al., 2009; Ligneul et al., 2018), these more naturalistic paradigms pose more difficulty in deconfounding factors of no interest, such as participants’ real-life knowledge and interests. Nevertheless, it would be important for future research to broaden the scope of application by, for instance, combining performance on this task with tracking sampling of real-life information (e.g., browsing news sites).

Furthermore, curiosity (Gruber et al., 2014; Kang et al., 2009; Kidd & Hayden, 2015) and choice (Cockburn et al., 2014; Murty et al., 2015) have both been shown to boost learning and memory, implicating couplings between reward-related areas (i.e., ventral striatum) and the hippocampus. However, the exact nature of the link between curiosity and choice in boosting memory remains unknown, an interesting question being whether enhanced memory for chosen items is mediated by increased curiosity. One step in that direction would be to introduce a curiosity assessment in paradigms that manipulate autonomy in learning—for example, memory for novel associations (Murty et al., 2015) or about stories and pseudoinformation—connecting these so far independently studied effects. Our findings may open the way for future work aimed at linking autonomy with curiosity and dopamine or noradrenaline as well as further investigating the nature of the relationship between choice, curiosity, learning, and memory.