Valuation of knowledge and ignorance in mesolimbic reward circuitry

Abstract

The pursuit of knowledge is a basic feature of human nature. However, in domains ranging from health to finance people sometimes choose to remain ignorant. Here, we show that valence is central to the process by which the human brain evaluates the opportunity to gain information, explaining why knowledge may not always be preferred. We reveal that the mesolimbic reward circuitry selectively treats the opportunity to gain knowledge about future favorable outcomes, but not unfavorable outcomes, as if it has positive utility. This neural coding predicts participants' tendency to choose knowledge about future desirable outcomes more often than undesirable ones, and to choose ignorance about future undesirable outcomes more often than desirable ones. Strikingly, participants are willing to pay both for knowledge and ignorance as a function of the expected valence of knowledge. The orbitofrontal cortex (OFC), however, responds to the opportunity to receive knowledge over ignorance regardless of the valence of the information. Connectivity between the OFC and mesolimbic circuitry could contribute to a general preference for knowledge that is also modulated by valence. Our findings characterize the importance of valence in information seeking and its underlying neural computation. This mechanism could lead to suboptimal behavior, such as when people reject medical screenings or monitor investments more during bull than bear markets.

Keywords: decision making; ignorance; information seeking; knowledge; valence.

Fig. 1.

Experimental design. (A and B) On each trial participants observed a pie representing the likelihood of (A) winning or (B) losing $1 on that trial. They also observed two offers (blue bars) representing the likelihood of the outcome being revealed (order of pie and bars counterbalanced across trials). Participants selected between these two offers, revealing their preference to observe the outcome cue on that trial. Then, either a knowledge cue (presented here in green) appeared, indicating that the informative outcome cue would be revealed in a few seconds (either win/zero/loss), or an ignorance cue (presented here in red) appeared, indicating that a noninformative outcome cue would be observed (XXXX), followed by a 1-s fixation cross. At the end of the study participants received the accumulation of gains/losses. (C) Postscan rating task in which participants indicated their desire to know following each pie, by moving a cursor on a scale from −300 (‟Not at all”) to +300 (‟Extremely”). Color associations were counterbalanced across individuals.

Fig. 2.

Preference for knowledge over ignorance is valence-dependent. Participants (A) selected the most informative offer more often during gain than loss trials and (B) rated their desire for knowledge higher during gain than loss trials. Participants (C) were more likely to choose the most informative offer and (D) rated their desire for knowledge higher on gain trials when they were more likely to win (orange curve) and on loss trials when they were less likely to lose (purple curve). Trendlines in C and D represent second-order polynomial fit (C: gain trials: R² = 0.962; loss trials: R² = 0.823; D: gain trials: R² = 0.969; loss trials: R² = 0.726). (E) Effect of lottery EV on knowledge preference, calculated for each individual participant as the slope between EV and the proportion of times they selected the most informative offer. (F) Uncertainty over outcomes following the delivery of informative or noninformative outcome cues, calculated for each trial as 0 when information is delivered and as the SD of the lottery when information is denied, then averaged separately for gain and loss trials across participants. Error bars represent ± 1 SEM. Two-tailed t test: *P < 0.05.

Fig. 3.

ROIs. ROIs were defined as 4-mm-radius spheres around the peaks of the Neurosynth ‟reward” map (threshold Z > 10). Specifically, based on a meta-analysis of 11,406 studies, this map reflects the relative selectivity with which voxels activate in relation to ‟reward,” by comparing all of the studies in the database that contained the term (671 studies for the term ‟reward”) to all those that did not. Three peaks were identified on the map, resulting in two ROIs shown in red: (A) bilateral NAc (peaks at [−10,8,−6] and [12,10,−8]) and (B) VTA/SN (peak at [4,−16,−12]).

Fig. 4.

Neural coding for the opportunity to obtain knowledge and remain ignorant is valence-dependent. (A) Individual participants’ parameter estimates of VD-IPE (equal to EV times IPE) during knowledge/ignorance cue in the VTA/SN ROI. This effect was significant at the group level, suggesting that IPEs are coded as a function of the EV of the outcome. (B) Individual components of the VD-IPE signal in VTA/SN showed positive tracking of the ‟actual opportunity to receive knowledge” component and negative tracking of the ‟expected opportunity to receive knowledge” component. (C) BOLD at the time of knowledge/ignorance cue was extracted for each trial and participant and analyzed in a mixed model with fixed and random effects of valence (binary coding: loss/gain), IPE, and their interaction. Resulting trendlines corresponding to fixed effect estimates of IPE tracking in VTA/SN are shown for gain and loss trials. (D) Individual differences in VD-IPE signal in the NAc ROI correlated with the effect of EV on information choice behavior as measured by the slope of the regression between EV and the proportion of trials in which knowledge is selected over ignorance. Choice of knowledge as a function of EV is plotted for an example participant.

Fig. 5.

Knowledge signal in OFC. Medial OFC cluster showing increased BOLD response during delivery of knowledge cues relative to ignorance cues (peak at [3,59,−5]; FWE whole-brain cluster-level corrected P < 0.05; threshold P < 0.001 uncorrected; overlaid on mean anatomical scan).

Fig. 6.

Factors influencing the amount paid for knowledge and ignorance. (A) Participants observed the evolution of a financial market after investing in two of its five companies then decided how much they were willing to pay to receive or avoid information about their portfolio value. (B) Example participant’s data showing a positive correlation between market change from the previous trial and signed WTP [coded positively when participants paid for knowledge and negatively when they paid for ignorance; R(200) = 0.401, P < 0.001]. Each dot represents one trial. (C) A mixed-effects model was run to predict signed WTP across all participants. Estimated coefficients, depicting the fixed effect of two significant factors, signed and absolute market change, are plotted. Additional control variables were added to the model and detailed in SI Appendix. (D) The effect of signed market change on signed WTP for knowledge over ignorance, extracted for each individual participant from the mixed-effects model. (E) Plotted is the difference in the total amount participants were willing to pay across all trials in which the market went up minus all trials in which the market went down, separately for when payment was for knowledge and ignorance. Error bars represent ± SE; *P < 0.05.