Aggression, a prevalent phenomenon across diverse populations, poses a significant risk for mental, physical, and social issues (Coccaro et al., 2017; Tuente et al., 2019; van den Berg et al., 2019). Consequently, there is a need to delve into effective prevention and intervention strategies. The SIP model encapsulates six sequential processes: encoding social cues, interpreting these cues, clarifying goals, generating responses, selecting and evaluating responses, and executing behavior. Notably, the second stage—interpretation—is often linked to reactive aggression, stemming from a hostile misinterpretation of ambiguous situations (Crick and Dodge, 1994). A pivotal element within this stage is the Hostile Attribution Bias (HAB), a cognitive predisposition that consistently interprets ambiguous situations as hostile rather than benign (Dodge and Coie, 1987). This bias can manifest in ambiguous contexts, such as an employee with a high HAB interpreting critical feedback from a manager as a personal attack, leading to aggressive responses. Conversely, an individual with a low HAB would likely perceive such feedback as constructive and refrain from aggression. Extensive research underscores the positive correlation between HAB and aggression, spanning adults, children, and adolescents (Crick and Dodge, 1996; Dodge and Somberg, 1987; Gagnon and Rochat, 2017; Xu et al., 2024). Encouragingly, the intensity of HAB is not fixed but malleable, as evidenced by CBM-I procedures, which have been shown to reduce aggression levels effectively (AlMoghrabi et al., 2018; Clarke et al., 2020; Vassilopoulos et al., 2015; Winnicki and Schmidt, 2023).

The classic CBM-I to address HAB involves presenting ambiguous scenarios in various formats, such as sentences, written vignettes, animated videos, or cartoon pictures, and requiring participants to choose between hostile and neutral interpretations, with feedback reinforcing the latter choice (AlMoghrabi et al., 2018; AlMoghrabi et al., 2024; Cougle et al., 2017; van Teffelen et al., 2021; Vassilopoulos et al., 2015). However, we believe that traditional paradigms have certain limitations. First, participants engage in a simple binary choice task that does not fully stimulate their cognitive abilities. Second, the training process involves providing correct feedback only after participants make a benevolent choice, which may lead them to infer the experimenter's intentions and deliberately select the benevolent option. Finally, some studies have indicated that using traditional paradigms for training has not effectively reduced individuals' HAB (Zeng et al., 2023).

The imagery-based Cognitive Bias Modification of Interpretation has emerged as a promising alternative to address these limitations. This approach, as demonstrated in Burnett Heyes et al. (2017) study, utilizes positive and mixed valence picture-word mental imagery training, where male adolescents were trained separately in field and observer perspectives. Positive imagery training led to significant improvements in mood and more favourable ratings of ambiguous pictures compared to mixed imagery training (Burnett Heyes et al., 2017). Furthermore, positive imagery generated from a field perspective effectively reduced negative interpretation bias in a language task. Schmidt and Vereenooghe (2021) successfully applied this imagery-based CBM-I method to HAB reduction (Schmidt and Vereenooghe, 2021). In Schmidt's study, ambiguous scenarios featuring adverse or potentially harmful outcomes were presented, with the intentions of other individuals causing these outcomes intentionally unclear. Participants were instructed to imagine themselves as the protagonist and were trained to generate positive interpretations from a benign perspective. This imagery-based training significantly reduced HAB and reactive aggression (Schmidt and Vereenooghe, 2021). Concerning these studies, we have adjusted the classic CBM-I (AlMoghrabi et al., 2024; Vassilopoulos et al., 2015) to promote active and spontaneous engagement in benign attributions. In our study, participants viewed ambiguous scenarios described in sentences and were asked to imagine themselves as the protagonist, experiencing the events from a positive perspective. They were then prompted to generate at least three positive interpretations for each scenario. We look forward to our training reducing HAB and aggression.

Despite limited research on the neural changes induced by HAB training through CBMI, previous studies have demonstrated the efficacy of CBM-I in exploring neural correlates across various cognitive domains. For example, the CBM has emerged as a therapeutic complement in the treatment of alcoholism, producing changes at behavioral and brain levels. Martínez-Maldonado et al. (2020) adopted CBM, in combination with the activation of alcohol-related memories, to explore the changes in the brains of patients with alcohol-use disorders. Their patients were divided into three groups: A-CBM (alcohol-related memory activation + CBM), N-CBM (neutral memory activation + CBM), and N-INT (no-intervention) groups. The results indicated that the CBM enhanced the functional connectivity of these brain regions in the resting state. There was a significant Group × Time interaction effect on the FC between the middle frontal gyrus (MFG) and the right middle frontal gyrus (MFG), as well as between the nucleus accumbens (NAcc) and the inferior frontal gyrus (IFG) (Martínez-Maldonado et al., 2020). Beevers et al. (2015) explored the effects of CBM in adult MDD patients. These patients were randomly assigned to 4-week CBM to reduce negative attention bias or 4-week placebo attention training. They found that the CBM increased the resting state functional connectivity between the middle frontal gyrus and dACC compared to the placebo. This circuit supports the control of emotional information (Beevers et al., 2015). Additionally, Li et al. (2016) studied the neural effects of a dot-probe-based CBM with positive, neutral, and negative faces in 41 young women with subthreshold depression (Li et al., 2016). Participants were randomly assigned to CBM, with 87.5 % training on positive cues (n = 24) and a placebo group receiving 50/50 negative/positive stimuli (n = 17). Resting state scans were conducted before and after training for both groups, and a non-depressed control group had scans before training only. Results showed decreased functional connectivity between the right insula and frontal-insular/supramarginal gyrus pre- and post-CBM, linked to improved depressive symptoms. To the best of our knowledge, in the context of CBM-I on HAB, no research has been conducted to examine its impact on the brain via the approach of resting-state functional connectivity. In light of the preceding findings, the second aim of the present study is to explore the neural mechanisms underlying CBM-I on HAB by utilizing resting-state functional connectivity.

While direct neurobiological evidence for CBM-I targeting Hostile Attribution Bias (HAB) remains limited, emerging studies have identified neural mechanisms associated with HAB and related psychological processes, offering critical insights into the neurocognitive underpinnings of HAB-focused training. Recent research has revealed that HAB is negatively correlated with functional connectivity strength between bilateral temporal poles (TP) and key regions of the Default Mode Network (DMN), including the middle temporal gyrus (MTG), medial prefrontal cortex (MPFC), and temporoparietal junction (TPJ) (Zhu et al., 2022). The DMN, predominantly active during rest, plays a central role in self-referential processing and social cognition (Andrews-Hanna et al., 2010; Tang et al., 2015). HAB is a modality for imputing intentions and a specific form of social extrapolation concerning the goals, beliefs, cognitive faculties, and conduct of others (Gagnon et al., 2024; Smeijers et al., 2019). Hence, CBM-I might be associated with an individual's default network. On the other hand, HAB requires inferring others' intentions, a process that shares neural substrates with Theory of Mind (ToM) – the ability to attribute mental states to others (Carlson et al., 2013; Zhu et al., 2022). The anterior medial prefrontal cortex (amPFC), a core hub of both the DMN and ToM networks (Friesen, 2021; Pardini et al., 2013), is critically involved in interpreting ambiguous social cues and regulating emotional responses during social reappraisal (Ochsner et al., 2012; Tan et al., 2022). Neuroimaging evidence suggests that hyperactivation of the amPFC occurs during ambiguous contexts of social decision-making and socially ambiguous laughter (volitional laughter), which has also been implicated in a mentalizing network. The amPFC is crucial for accessing emotional and social knowledge related to oneself or others, especially during mentalizing or introspective processes (Corcos et al., 2012; McGettigan et al., 2015). It also plays a key role in appraisal processes that help predict social and emotional outcomes, not just for current social events but also based on memories of past experiences (Vrtička et al., 2009). Taken together, we hypothesize that CBM-I would modulate the activity of brain regions associated with the DMN, particularly in the amPFC, to facilitate cognitive restructuring of hostile schemas.

In summary, our research has made adjustments to the existing CBMI, and we aim to explore its impact on HAB and reactive aggression. Additionally, we seek to investigate the neural basis underlying this training further by leveraging resting-state functional connectivity. Based on this, we have put forward the following hypotheses:

Hypothesis 1

Compared to the pre-test, the training group will significantly reduce HAB and reactive aggression after undergoing the CBMI, while the control group will show no significant changes in hostile attribution bias or reactive aggression.

Hypothesis 2

After the intervention, the training group will experience a decrease in both HAB and reactive aggression compared with the control group.

Hypothesis 3

The changes in hostile attribution within the training group might be associated with the DMN. In contrast, no significant changes will be observed in the brains of the control group.