Background
Decision-making is important in daily life and has significant consequences. As human beings, since we live in a complex social world, our decision-making during interaction is not only shaped by our individual goals but also by those of others’ [
1,
2]. Decision making in the context of others is termed social decision-making, or more specifically, strategic interactive decision-making. Paradigms from game theory, which have interactive characters, have been increasingly used to investigate social decision-making [
1]. Using these simulating games of interpersonal and group interactions, abnormal social decision-making behaviors have been reported in patients with schizophrenia, such as less trust in investment with another counterpart [
3], non-strategically less free riding in public goods game [
4], and less rejections of unfair offers when splitting a sum of money with a counterpart [
5,
6]. However, the underpinnings of these deviant behaviors in schizophrenia are still unclear.
Previous studies have implicated that an ability termed Theory of Mind (ToM) may be a candidate mechanism of social decision-making [
1,
7]. ToM refers to the cognitive ability to represent one’s own and others’ mental states to further explain and predict behaviors [
8]. This ability is often assumed to be involved in social decision-making, in which inferring intentions of others involved in the task is implicitly required. We can take the ultimatum game (UG) as an example. During this game, two players (proposer and responder) obtain a sum of money together. The proposer first specifies how to split the money between the two of them, and then the responder makes a decision to accept or reject the offer. If the offer is accepted, the two players will get their own share; if it is rejected, neither of them receives anything [
9]. Researchers observed that healthy participants as responders showed both behavioral and neural differences between the situations in which, respectively, the proposer is a human with intentions or a computer [
10,
11]. Thus, the potential involvement of ToM has been suggested. Specifically, in behavioral studies, rejection rates were higher when the unfair offers were proposed by the human being rather than the computer [
11]. Therefore, based on the assumption that it is not necessary to consider the agent’s intention in the computer condition with the other parameters being the same, the differences in the rejection rates reveal the different involvement of ToM in the two conditions. Furthermore, neuroimaging studies found that compared to playing with the computer, playing with the human being produced stronger activation in brain regions which are overlapped with the neural networks of ToM [
12], such as the anterior paracingulate cortex and the posterior superior temporal sulcus [
10]. Overall, these findings suggest an involvement of ToM in the responder’s behavior in the UG. In contrast to these theoretical predictions and discussion on the potential role of ToM in social decision-making [
1,
4,
13,
14], no study has directly measured the relationship between social decision-making and ToM in patients with schizophrenia, in which ToM deficits have been consistently reported [
12,
15‐
19].
The current study aimed to explore the potential relationship between ToM and social decision-making during the mini ultimatum game (mini UG) in patients with schizophrenia. Distinct from the abovementioned classic UG, in the mini UG, the proposer is given two options to choose between on each occasion. One is always 8 vs. 2 (the proposer gets 8 and the responder gets 2) paired with one of four possible alternatives: 5 vs. 5, 2 vs. 8, 8 vs. 2, and 10 vs. 0 [20]. In addition to the chosen option, which corresponds to the proposal in the UG, the additional unchosen alternative in the mini UG can provide clues to the responder to infer intentions underlying the chosen option by the proposer. Consequently, the responder will have different rejection rates to the same chosen option as the alternative changes [
20]. Thus, compared to the UG, the modification of the mini UG directly expose the underlying intentions of the proposer’s choices [
11,
20‐
24]. The modification also makes it possible to test whether the unfairness itself or the underlying intentions of the offers drive the responder to make a decision. Therefore, we reasoned that the mini UG should be an efficient paradigm to explore the role of ToM in social decision-making.
In the current study, we first compared the social decision-making behaviors as the responder during the mini-UG in patients with schizophrenia and healthy controls. Next we explored the influence of ToM on responder’s choice during the mini UG. We speculated that compared to the healthy controls, patient with schizophrenia may accept more disadvantageous offers but reject more advantageous offers, based on previous studies, in which the patients with schizophrenia often showed less rejection rates to the unfair (disadvantageous) offers in the classic UG [
5,
6] but higher rejection rates to the fair (advantageous) offers in the classic UG [
5] or in the mini UG [
21]. More importantly, we hypothesized that except in the condition in which the proposer has no alternative (both options are 8 vs. 2), the abnormal behaviors of patients in the mini UG would correlate with their ToM deficits, as they may have difficulty in inferring the intentions of the proposer given the unchosen options. To test this possibility, we used the Theory of Mind Picture Stories Task (TMPST) [
25] to measure ToM and examined the mediation effect of ToM in the mini UG. We also noted that neurocognition deficits are well-established in schizophrenia and have a close relationship with social cognition, including ToM [
26,
27]. Furthermore, specific aspects of neurocognition, such as working memory and executive function, are closely related to decision-making behaviors [
28,
29]. Therefore, in the present study, in addition to the focus on the role of ToM in social decision-making, we also explored the mediation effect of neurocognition on social decision-making.
Discussion
In the current study, we found that the patients showed lower rejection rates to the disadvantageous offers and higher rejection rates to the advantageous offers, and reduced sensitivity to the fairness-related context changes in the mini UG. These findings validated the previous findings [
5,
6,
22,
40]. The new contribution of the current work is that we directly investigated the roles of ToM and neurocognition deficits in impaired social decision-making of patients with schizophrenia during the mini UG. We found that the ToM deficits but not dysfunctions in neurocognition, partly mediated the significant group differences in the mini UG during which the intentions of the proposer needed to be inferred.
In the mini UG, an interactive game related to fairness principle in social life, the patients with schizophrenia showed abnormal response patterns compared to the healthy controls. On one hand, patients rejected less disadvantageous offers. Though these offers were in different contexts, all of them were unfair to the participants who played as the responder in this game. This finding echoed the previous observation of less rejections of unfair offers in the classic UG [
5,
6] and was also compatible with the finding of more acceptances of unfair offers from the human proposer in populations with schizotypal symptoms [
40]. On the other hand, we observed that patients with schizophrenia showed more rejections to the advantageous offers. Higher rejections were found both to the fair and hyper-fair offers, but on the last one group differences achieved a significant level. Still, this finding corroborated the results of previous research using both UG and mini UG that more fair offers were rejected by patients with schizophrenia [
5,
22]. In brief, using the mini UG we can detect the abnormal social decision-making behavior in patients with schizophrenia.
Furthermore, we found that the patients showed reduced sensitivity to the change of fairness-related contexts, which suggested that the ToM deficits may be a potential psychopathological mechanism for the altered rejection rates observed in the patients with schizophrenia. In order to directly test the possibility that the ToM deficits may partly explain the abnormal social decision-making in schizophrenia, we analyzed the mediation effect of ToM in the group differences found in the mini UG. We found that the patients with schizophrenia performed worse on the questionnaire scores in the TMPST, suggesting that patients were worse at attributing cognitive mental states (e.g., belief and intentions) of characters in the stories. This finding echoed the consistent observation of ToM deficits in schizophrenia in the previous studies [
15,
17,
41]. In conformity with our hypothesis, the questionnaire scores mediated the group differences in rejections both to the disadvantageous and advantageous offers in the conditions when another alternative was available and intentions were exposed, but not in the case when both options were the same 8 vs. 2. Specifically, rejection rates in the three disadvantageous conditions (hyper-unfair offer, hyper-fair alternative, and fair alternative) positively correlated with the TMPST questionnaire scores. The correlations suggested that given the alternative options, the participants with better ToM ability may detect the intentions underlying the choices of the proposer and thus were more likely to reject the intentionally unfair offers. On the other hand, the rejection rates in the advantageous condition (hyper-fair offer) negatively correlated with the TMPST questionnaire scores. This result indicated that the participants with higher ToM abilities appreciated the generous intentions and thus were less likely to reject the hyper-fair offers. Interestingly, Sally and Hill reported that children with autism spectrum disorder (ASD) who had ToM deficits showed similar behavioral pattern in the UG [
42], but no other related studies have been found so far. In conclusion, ToM deficits in patients with schizophrenia impair the social strategic ability and make it difficult for the patients to fully integrate intention inferring into their decision-making in the social contexts; thus ToM deficits can partly account for the abnormal social decision-making during the mini UG.
In addition to ToM, we also investigated the role of neurocognition in the social decision-making. Consistent with neurocognition dysfunctions found in schizophrenia [
43], the patients in present study acquired significantly less composite scores of BACS. Significant group differences existed on four out of six subtests. In a previous study, researchers suggested that patients with schizophrenia have difficulties in maintaining the representation of the goal and integrating it within the complex context (e.g., the mini UG task) due to their impairments in working memory and proactive control [
44]. However, in the present study, we found that the relationships between BACS and performances of patients in the mini UG were weak and mediation analyses showed that the neurocognition (BACS) alone did not have a mediation effect in the group differences. Only considering the working memory measurement (subtest scores of digit sequencing), we also did not find mediation effect. Similar to our case, in a previous study with UG, Csukly et al. also measured participants’ neurocognition including verbal learning and memory, executive control and working memory; they reported that adding these neurocognition measures as covariates into analyses did not affect their results that patients with schizophrenia rejected less unfair offers and more fair offers [
5]. No more research was found which directly studied the influences of neurocognition in abnormal social decision-making in patients with schizophrenia. Therefore, more investigations are still needed to further explore the specific relationships between social decision-making and neurocognition. In the present mini UG task, the abnormal behaviors exhibited in the patients with schizophrenia were more related to their impaired ToM rather than their neurocognition deficits.
There are several limitations of this study that need to be addressed. First, in the present study, we did not record the parental level of educational and social economic status in the participants; thus, the influence of parental background on the social decision-making cannot be excluded. Second, the current study did not assess the potential effects of clinical heterogeneity or specific clinical symptoms on social decision-making behavior in patients with schizophrenia, as the relatively small sample size in the current study was inappropriate for categorizing the participants based on their clinical characteristics. Future investigations with subgroups of patients are needed to test whether the abnormal behavioral pattern and its potential mechanism are consistent across these subgroups. For example, it is possible that the decisions made by patients with persecutory delusion may be different from those made by patients with negative symptoms and the underpinnings of social decision-making behavior may also be different, considering that the two subgroups may perform differently on ToM tasks [
45‐
47]. Third, the potential role of emotional ToM ability in social decision-making also requires further exploration, as the task used in the present study focused mainly on the cognitive aspect of ToM. Finally, though the mediation effect of ToM in social decision-making deficits in schizophrenia was identified, the question of what exact aspect of ToM led to the deviant performances in the mini UG cannot be answered based on the current findings. Future studies can address this question using a computational model that can fit behavioral data, i.e., trial-by-trial decisions of patients and healthy controls. Clearly, to parameterize inter-subject variability in a normative or formal sense, a normative model is needed. This could be addressed using the hierarchical Gaussian filter [
48] or by active inference for Markov decision processes [
49]. The second method may be more suited to the mini UG type of game, and we are currently developing and validating these models for future applications in performance analyses of games, including the mini UG.
Acknowledgments
The authors gratefully acknowledge Chris Frith for his comments and suggestions on an earlier version of this manuscript. The authors acknowledge the three reviewers for their critical review of the manuscript and for suggesting substantial improvements. The authors also thank Karl J Friston for his helpful comments during revision and thank Andreas Hula for his help on the binary logistic regression modelling.