How expert advice influences decision making

Abstract

People often use expert advice when making decisions in our society, but how we are influenced by this advice has yet to be understood. To address this, using functional magnetic resonance imaging, we provided expert and novice advice to participants during an estimation task. Participants reported that they valued expert advice more than novice advice, and activity in the ventral striatum correlated with this valuation, even before decisions with the advice were made. When using advice, participants compared their initial opinion to their advisor's opinion. This comparison, termed the "opinion difference", influenced advice utilization and was represented in reward-sensitive brain regions. Finally, the left lateral orbitofrontal cortex integrated both the size of the opinion difference and the advisor's level of expertise, and average activity in this area correlated with mean advice utilization across participants. Taken together, these findings provide neural evidence for how advice engenders behavioral change during the decision-making process.

Figure 1. Experimental task.

Participants had 8.5 s to estimate the real monthly rental price of an apartment in New York City by using four attributes (square meters, number of rooms, number of bathrooms, neighborhood). They then discovered the expertise level of their advisor (Time 1; 2.5 s). After a short delay, participants were then given advice on the rental price and allowed to adjust their answer (Time 2; 7 s).

Figure 2. Behavioral data illustrating the utilization of advice.

(A) Participants used expert advice more than novice advice (p<0.001). Important to note, all advice that participants received, from both experts and novices, was the actual price of the apartment (see Materials and Methods). Participants also used advice significantly more when the advice amount was close to their first estimate (low opinion difference) compared to when the advice was far from their first estimate (high opinion difference) (p = 0.017). (B) Histogram of individual differences in usage of advice with respect to expertise (mean expert WOA minus mean novice WOA). Participants demonstrated variability in their usage of advice from different sources. Notably however, all of the participants used expert advice qualitatively more than novice advice (no participants below zero). (C) Participants exhibited shorter reaction times when using expert advice than when using novice advice (p<0.001). They also responded more quickly when the opinion difference was high compared to when the opinion difference was low (p = 0.039). (D) After the experiment, participants rated the expert advice as being more valuable than the novice advice (scale from 1 =  low to 5 =  high; p<0.001). Error bars represent standard error of the mean. All analyses performed using repeated measures ANOVA unless otherwise indicated. OD  =  opinion difference.

Figure 3. Brain regions showing a main effect between the expert and novice condition.

(A) When contrasting expert > novice at Time 1, participants showed greater changes in BOLD signal in the ventral striatum upon discovering that their advisor will be an expert compared to discovering that their advisor will be a novice. (B) When contrasting expert > novice at Time 2, participants showed greater changes in BOLD signal in the medial prefrontal cortex when using advice from experts compared to using advice from novices. (C) Parameter estimates in the medial prefrontal cortex at Time 2. Bars indicate standard error of the mean. BOLD activation maps thresholded at Z >2.3, p<0.05, cluster corrected. L  =  left, OD  =  opinion difference, PFC  =  prefrontal cortex.

Figure 4. Brain regions demonstrating differential activity due to size of opinion difference upon revealing the advice (Time 2).

(A) When contrasting high opinion difference trials > low opinion difference trials, changes in BOLD signal were greater in the lateral orbitofrontal cortex and the ventromedial prefrontal cortex. (C) When contrasting low opinion difference trials > high opinion difference trials, changes in BOLD signal were greater in the ventral striatum and the anterior cingulate cortex. (B,D) Parameter estimates in these regions. Bars indicate standard error of the mean. BOLD activation maps significant at p<0.01, FDR corrected within a priori defined areas (see Materials and Methods). L  =  left, OD  =  opinion difference, OFC  =  orbitofrontal cortex.

Figure 5. Interaction effect between expertise and size of opinion difference when advice was revealed (Time 2) in left lateral orbitofrontal cortex.

(A) BOLD activity in an interaction contrast ((expert high opinion difference trials > expert low opinion difference trials) > (novice high opinion difference trials > novice low opinion difference trials)) revealed a significant interaction effect in the left lateral orbitofrontal cortex. BOLD activation map is significant at p<0.01, FDR corrected within a priori defined areas (see Materials and Methods). (B) Parameter estimates in the left lateral orbitofrontal cortex. Bars indicate standard error of the mean. (C) The mean activation across all advice trials in this lateral orbitofrontal region predicts the mean weight of advice over all advice trials for individual participants (Pearson’s r = 0.488, p = 0.007). L  =  left, OD  =  opinion difference.