Financial incentives facilitate the neural computation of prosocial decisions stronger in low empathic individuals

Financial incentives are commonly used to motivate behaviours. There is also evidence that incentives can decline the behaviour they are supposed to foster, for example, documented by a decrease in blood donations if a financial incentive is offered. Based on these findings, previous studies assumed that prosocial motivation is shaped by incentives. However, so far, there is no direct evidence showing an interaction between financial incentives and a specific prosocial motive. Combining drift-diffusion modelling and fMRI, we investigated the effect of financial incentives on empathy, i.e., one of the key motives driving prosocial decisions. In the empathy-alone condition, participants made prosocial decisions based on empathy, in the empathy-bonus condition, they were offered a financial bonus for prosocial decisions, in addition to empathy induction. On average, the bonus enhanced the information accumulation in empathy-based decision. On the neural level, this enhancement was related to the anterior insula, the same region that also correlated with empathy ratings. Moreover, the effect of the financial incentive on anterior insula activation was stronger the lower a person scored on empathy. These findings show that financial incentives enhance prosocial motivation in the absence of empathy but have little effect on high empathic individuals.

prosocial behaviours (Balliet et al., 2011;Stoop et al., 2018). For example, in a meta-analysis, 46 Balliet and colleagues found that reward positively affects cooperation (Balliet et al., 2011). 47 Consequently, financial incentives could increase the motivation to behave prosocially (Ariely et  blood if they are paid to do so, compared to the amount of blood that they donate without 53 payment, i.e., to help others (Titmuss, 1970;Niza et al., 2013). In line with these observations, 54 other studies have shown that adding financial incentives can reduce prosocial behaviours 55 (Bowles, 2008;Ariely et al., 2009;Holmås et al., 2010). In sum, the evidence regarding the 56 effects of incentives on prosocial decisions is inconsistent, and mainly based on behavioral 57 observations that do not provide insights in the underlying motivational processes. As a result, it 58 remained unclear whether and how financial incentives interact with a specific prosocial motive. 59 Overcoming this limitation, our study directly investigated how a financial incentive 60 shapes prosocial decisions that are driven by a specific prosocial motive, i.e., empathy.

61
Incorporating previous approaches, we used a well-established decisions task (i.e., a modified 62 version of a binary dictator game (Hein et al., 2016b)). Extending previous studies, we activated 63 a specific prosocial motive (empathy) before participants entered the decision task, and, in one 64 condition, added a financial incentive. This allowed us to investigate how financial incentives 65 change the processing of prosocial decisions that are driven by one specific, carefully controlled 66 motive. To control for other motivations that might play a role besides empathy (self-image 67 concerns; reciprocity), the incentive was offered in private, the decisions were kept anonymous, 68 and the participants knew that they would not meet the other players after the study. This measure 69 is important because it minimizes participants' motivation to maintain a positive public image, 70 i.e., a different motive that may affect participants' prosocial decisions besides empathy (Benabou the individual strength of the empathy motive and the propensity to act prosocially, e.g., 80 decisions that maximize the outcome of another person at costs to oneself (Batson et al., 1995; 81 Decety et al., 2016). The stronger the empathy motive, the stronger the propensity to decide in 82 favour of the other person. Previous social psychology work has investigated how empathy is 83 shaped by selfish motives, such as the motive to withdraw from a stress-inducing situation 84 (Batson et al., 1981). However, to the best of our knowledge, there are no previous studies that 85 tested how financial incentives affect the components of empathy-based prosocial decisions. 86 The study consisted of two parts (Fig. 1). In part 1, the empathy motive was activated 87 towards one partner (a confederate). In the following allocation task, participants allocated points 88 to the respective partner (here driven by empathy; empathy-alone condition). Next, the 89 confederate was replaced by a new individual that served as a partner for part 2. In part 2, the 90 empathy motive was activated again. However, before starting the decision task, the participant 91 was told that she would receive a bonus if she decided prosocially in the majority of the decision 92 trials. In the following allocation task, participants again allocated points to the respective partner 93 (here driven by empathy and the financial incentive; empathy-bonus condition). The order of the 94 two conditions (empathy-alone and empathy-bonus) was counterbalanced across participants and 95 the two confederates. 96 97 Fig. 1 Overview of an exemplary experimental procedure. The study consisted of two parts. In this example, in part 98 1, the empathy motive was activated towards one confederate (the empathy-alone partner). In the following 99 allocation task, participants allocated points to the empathy partner (i.e., driven by the empathy motive). Next, the 100 confederate was replaced by a new individual that served as partner for part 2. Again, the empathy motive was 101 activated. After the empathy motive induction additionally a bonus for choosing the prosocial option in the majority 102 of trials in the subsequent allocation task was offered (empathy-bonus partner). Thus, in the following allocation 103 task, participants allocated points towards the empathy-bonus partner (i.e., driven by the empathy motive and the 104 additionally offered bonus). The order of motive induction (empathy-alone, empathy-bonus) was counterbalanced 105 across participants and both confederates. The respective partner was indicated by a cue in one of two 106 counterbalanced colors. person in pain ( Fig. 2A). To allow participants to simulate the state (pain) of the other person, in 113 some trials, participants received painful stimulation themselves.

114
During the allocation task inside the fMRI scanner, participants allocated points to the 115 partners at a cost to themselves (Fig. 2B). The allocation of points towards the one partner

143 144
To specify how incentives modulate empathy-related decisions, we used drift-diffusion   One assumption is that financial incentives may enhance empathy-related prosocial

237
The study started with the empathy induction, followed by the allocation task towards the 238 first confederate. After replacing this confederate, the same procedure (empathy induction 239 followed by the allocation task) was repeated with the second confederate ( Fig. 1). In the 240 empathy-alone condition, the allocation task started immediately after the empathy induction. In 241 the empathy-bonus condition, after the empathy induction, participants were told that they would In each empathy-induction trial, first we presented a coloured arrow indicating the person 273 who will receive the following electric stimulation for 1000 ms. After this cue, a fixation cross 274 was presented for 1000 ms, followed by a coloured lightning bolt shown for 2000 ms.

275
Participants were informed that a blinking dark-coloured lightning bolt indicates a painful 276 stimulus, whereas a blinking light-coloured lightning bolt indicates a non-painful stimulus. After 277 receiving or observing the electric stimulation, we showed a 9-point rating scale with the question 278 "How do you feel?". The scale ranged from -4 (labeled "very bad") to +4 (labeled "very good").

279
Participants had to respond within 4000 ms ( Fig. 2A). The empathy induction consisted of 30 280 trials: 10 that were ostensibly painful for the partner (other-pain trials), 5 that were not painful for 281 the partner (other-no-pain trials), 5 painful trials for the participant (self-pain trials), and 10 non-282 painful trials (self-no-pain trials) for the participant. The self-pain trials were added to allow 283 participants to simulate the state (pain) of the other person. To test their potential influence on 284 empathy changes, we compared the ratings in other-pain trials that were preceded by a self-pain 285 trial (i.e., empathy ratings under the condition of self-pain experience) with the ratings in other-286 pain trials that were preceded by an other-pain trial (i.e., empathy ratings without preceding self-287 pain experience). The results showed no difference between the other-pain ratings after self-pain 288 and the other-pain ratings without prior self-pain (t(61) = 0.34, p = .73). Based on these results, the 289 self-pain experience had no significant effect on empathy changes during empathy induction. were allocated with a ratio of 90% (self) to 10% (other). Participants' losses were symmetrical to 308 the partner's gains. For example, a total of 1000 points were distributed with self:other ratios of 309 60:40 (600:400 points), 90:10 (900:100 points). Thus, the participant's loss is 900 -600 = 300 310 points, which corresponds exactly to the gain of the partner (400 -100 = 300 points). We used 311 these fixed and symmetrical ratios to minimize unspecific effects of loss aversion.

312
Each decision-trial started with an inter-trial interval indicated by a fixation cross 313 presented for a period jittered between 4000 and 6000 ms (Fig. 2B). After this, participants saw attaching the electrodes at the index finger of the right hand and connecting them to the single-328 current stimulator, the respective person was asked to press the button for defining the current 329 threshold and to decide when she is feeling the stimulationthe value of this threshold was used 330 as painless stimulation. In a second run the participant was asked to press the same button, but 331 now to hold it pressed until the pain was at an unacceptable level and then to releasethis 332 threshold was used for the painful stimulation.

333
Experimental design and statistical analyses 334 The aim of our study was to compare prosocial decisions driven by empathy alone with

346
Empathy ratings showed a right-skewed distribution (Shapiro-Wilk W = .94, p < .01), so 347 the data was log-transformed to normal distribution. Pearson correlation was computed between 348 the empathy ratings and the empathic concern scale (EC) from the Interpersonal Reactivity Index 349 (IRI) (Davis, 1980). In further data analyses, we used linear models within condition and linear 350 mixed models (lmm) with participants as random effect between conditions.

454
Empathy was induced with comparable strength in both conditions.

455
To quantify the strength of the induced empathy, we calculated the participants' trial-by-

466
According to these results, the induced motive is related to empathic concern rather than personal 467 distress.

468
The financial incentive increased the frequency of prosocial decisions, in particular if empathy 469 was low. alone condition (Fig. 3B).

522
We compared the speed of information accumulation (drift rate; v-parameters), the initial

565
The higher the speed of information accumulation in the empathy-bonus condition, the stronger the neural response   Table 1. Neural results of the second-level regression between prosocial decision-related activity in the Empathy-575 bonus condition and the speed of information accumulation (v-parameter) in the Empathy-bonus condition. The 576 asterisk (*) indicates activations that are significant at 5% FWE voxel-based inference. We also conducted 577 explorative analyses with 5% FWE cluster-based inference with a cluster-forming threshold of P < .001and a 578 minimal cluster size of k = 50. Please note that peak-coordinates derived from cluster-wise inference only provide 579 information about activated brain components, but not the exact brain region (Woo et  if the observed AI region (i.e., the region that correlated with the speed of information 585 accumulation in the empathy-bonus condition) was also related to the empathy ratings that we 586 collected prior to the allocation task. To do so, we extracted the average of the beta estimates conditions, we can infer that the observed AI activation, in general processes individual 592 differences in empathy, i.e., unbiased by the specific experimental conditions. The higher a 593 participant's empathy ratings, the stronger the neural response in the AI region, i.e. the same 594 region that correlated with the speed of information accumulation in the empathy-bonus 595 condition (Fig. 4C). 596 The financial incentive has a differential effect on anterior insular activation in high and low 597 empathic individuals.

598
Given that the v-parameter and empathy ratings both are processed in the same AI region,

599
it is plausible to assume that the two variables interact. To test that we conducted a linear mixed  Table 1 shows that also other brain regions correlated with the individual increase in v-  conditions. The beta estimates reflect the average of AI activation from the empathy-bonus and the 668 empathy-alone condition, extracted from the same AI clusters that correlated with the v-parameter in the 669 empathy-bonus condition (shown in Fig. 4B) To test the robustness of the differential effects in the empathy-bonus and the empathy-  (Fig. 3A), in particular in individuals that scored low on empathy (Fig. 3B). 693 The finding that the financial bonus enhanced the frequency of prosocial decisions is in line with bonus compared to the empathy-alone condition (Fig. 4A). In contrast, the incentive had no 702 significant effect on participants' initial prosocial preferences, i.e., the preference of making a 703 selfish or prosocial decision with which they entered the decision process (captured by the z-704 parameter).

705
Outside the domain of prosocial decisions, there is evidence that the efficiency of (captured by an increased speed of information accumulation) is related to an increase of AI 720 responses (Fig. 4B) and that this same AI region also correlated with individual differences in 721 empathy (Fig. 4C). 722 Adding a novel aspect, our findings reveal how financial incentives alter the effect of 723 empathy on the computation of prosocial decisions in the anterior insular cortex. After offering a 724 bonus in the empathy-bonus condition, the relationship between empathy ratings and drift rate 725 and empathy ratings and AI estimates was no longer significant, indicating that in the presence of 726 an incentive, empathy was no longer a significant driver of prosocial decisions. Interestingly, the 727 interaction between the empathy ratings and the drift rate significantly reduced AI activation in 728 the empathy-bonus condition (Fig. 5A) while increasing it in the empathy-alone condition (Fig.   729   5B). This indicates that in the empathy-bonus condition, the strength of the empathy motive  In our study, empathy was conceptualized as a motive that can drive prosocial decisions.

737
And indeed, the empathy ratings of our participants that correlated with empathic concern (but 738 not personal distress) facilitated the prosocial decision process in the empathy-alone condition, in 739 line with previous findings (Batson et al., 1995;Decety et al., 2016). That said, the result that 740 financial incentives counteracted the facilitating effect of empathy on prosocial decisions in 741 highly empathic individuals might indicate that highly empathic individuals are less motivated to 742 empathize in the presence of an incentive, an assumption that supports the notion that empathy 743 itself is a motivated state (Zaki, 2014).

744
The financial incentive for prosocial decisions was offered in private, and self-image 745 concerns were reduced as far as possible, at least with regard to public reputation. However, 746 some highly empathic participants nevertheless showed an incentive-related decline in prosocial 747 decisions (see also Fig. 3B). It is conceivable that highly empathic participants feel insulted by 748 the bonus because "being paid to be nice" undermined their intrinsic empathic motivation that 749 otherwise (i.e., in the empathy-alone condition) drives their prosocial decisions. Thus, although 750 on average our findings show that the incentive increased the frequency of prosocial decisions 751 compared to an empathy-alone condition, it is still possible that it undermines prosocial behavior 752 in highly empathic participants. To test this assumption, future studies should test the effect of 753 financial incentives on empathy-based decisions in extreme groups, i.e., groups of extremely high 754 or low empathic individuals. Moreover, it would be interesting to use a trial-by-trial bonus 755 manipulation that allows for modelling the effect directly as part of the DDM.

756
In summary, our current results indicate that financial incentives offered in private