Can monetary incentives overturn fairness-based decisions?

Abstract

Fairness norms and resulting behaviours are an important prerequisite for cooperation in human societies. At the same time, financial incentives are commonly used to motivate social behaviours, yet it remains unclear how financial incentives affect fairness-based behaviours. Combining a decision paradigm from behavioural economics with hierarchical drift-diffusion modelling, we investigated the effect of different financial incentives on two types of fairness-based decisions in four experimental groups. In two groups, participants divided points between themselves and a disadvantaged person, inciting fairness-based compensation behaviour, in two other groups they divided points between themselves and a fairness violator, inciting fairness-based punishment behaviour. In addition, each group received financial incentives that were either aligned or in conflict with the respective fairness-based behaviour. This design allowed us to directly investigate how different incentives shape the cognitive mechanism of fairness-based decisions and whether these effects are comparable across different fairness domains (fairness-based punishment versus fairness-based compensation). Results showed that offering conflicting incentives diminished fairness-congruent decision behaviour and rendered the fairness-congruent decision process less efficient. These findings demonstrate that financial incentives can undermine fairness-based behaviour, and thus are relevant for the development of incentive schemes aimed at fostering cooperative behaviour.

Keywords: compensation; drift-diffusion modelling; fairness; financial incentives; punishment.

Figure 1.

Schematic illustration of the DDM. The DDM conceptualizes the decision-making process as a noisy accumulation of information (blue line). This process can be characterized in terms of the speed of information uptake (v-parameter), the initial bias towards one of the decision options (z-parameter), and the amount of information required until a decision is reached (a parameter). Please note that values of the drift rate above the zero point of the accumulation process are positive, i.e. with a drift towards the upper boundary (here, the fairness-congruent option, i.e. decisions in favour of the other in the compensation groups and decisions in favour of the self in the punishment groups), and values below zero are negative [47], i.e. with a drift towards the lower boundary (here, the fairness-incongruent option, i.e. decisions in favour of the self in the compensation groups and decisions in favour of the other in the punishment groups).

Figure 2.

The trial procedure of the observation phase (top), example trials of the decision task for all groups (bottom), and the different experimental conditions (types of the receiver) in the punishment groups (blue arrow) and the compensation groups (orange arrow). Each trial of the observation phase (top) started with a fixation period. Then, participants observed allocation decisions in a binary dictator game, most of which were unfair (80% unfair, 20% fair). In the example trial shown here, the allocator assigned 102 points to the receiver and 498 points to the self. The decision of the allocator was highlighted by a green rectangle. After having observed the decision of the allocator, participants were asked to rate the fairness of the decisions on a 9-point scale ranging from very unfair (−4) to very fair (+4). In the decision task (bottom), participants were the allocators. In the two punishment groups, they divided points between themselves and the allocator of the previous observation phase (left panel) or to self and a previously uninvolved player (baseline partner; middle panel). In the two compensation groups, participants divided points between themselves and the receiver of the previous observation phase (right panel) or to the self and a previously uninvolved player (baseline partner; middle panel). The type of receiver (allocator of the observation phase versus uninvolved player; receiver of the observation phase versus uninvolved player) was indicated by a cue following the fixation period. In parallel, the incentive was shown (varying between 0 and 500 points). Depending on the experimental condition (table 1), the incentive was provided for decisions in favour of the self or in favour of the other. Next, participants were presented with the decision options, and allocated points in favour of the self or in favour of the respective other (allocator/receiver of the observation phase; uninvolved player). The decision was highlighted by a green rectangle.

Figure 3.

Exemplary experimental procedure. Each participant was assigned to one of four groups. During the observation phase, in which the participant was merely in the observer role, the fairness motive was associated with two other players (the allocator or the receiver). In the following decision task, participants were in the role of the allocator and divided points between themselves and the previously observed fairness partner (allocator or receiver) as well as themselves and a baseline partner (previously uninvolved player). Depending on the group assignment, participants were offered trial-by-trial varying incentives for making decisions in favour of the self, or for making decisions in favour of the other. For details regarding each task, please refer to figure 2.

Figure 4.

Overview of the reward structure. (a) In the two groups, in which decisions in favour of the other are incentivized (conflict punishment group, aligned compensation group), participants received their pay out based on the points of the chosen distribution. Additionally, they received the pay out based on the trial-specific incentive for the drawn trial(s) in which they chose in favour of the other. (b) In the two groups, in which decisions in favour of the self were incentivized (aligned punishment group, conflict compensation group), participants received their pay out based on the points of the chosen distribution. Additionally, they received the pay out based on the trial-specific incentive for the drawn trial(s) in which they chose in favour of the self.

Figure 5.

Effect of conflict degree and incentive level in the two fairness domains. Conflicting incentives diminish the frequency of fairness-congruent responses, i.e. decrease the frequency of decisions in favour of the other in the compensation domain and the frequency of decisions in favour of the self in the punishment domain. This effect is the larger, the higher the incentive level.

Figure 6.

Results of the exploratory DDM analyses. (a) Posterior distribution of the regression weights indicating the influence of incentive level on the z-parameter. Less reach of the distribution beyond the 0-value corresponds to a larger effect. (b) Posterior distribution of the v-parameter. Less overlap between the orange and light blue (compensation domain), and the dark green and dark red (punishment domain) distribution, corresponds to a larger effect of conflicting incentives on the respective fairness-based behaviour.

Figure 7.

Posterior distributions of the estimated v-parameter in the four groups. Conflicting incentives (blue) markedly reduce the v-parameter as compared with aligned incentives (red). Generally, the decision process in favour of the self (punishment groups, darker shades) is more efficient than the decision process in favour of the other (compensation groups, lighter shades).