Integrating information is particularly crucial when decisions contain complex information. In decision neuroscience, it has been widely studied how overall decision value is encoded in the brain, but how information is integrated prior to encoding overall value remains largely unclear. Recently, our neuroimaging study has shown that the lateral frontopolar cortex (FPl), a uniquely well-developed region in the human brain, engaged specifically in complex choice but not in simple choice (Law et al., 2023). Aided by an artificial neural network, we found that FPl involved a parallel processing to extract and integrate the complex choice information for decision making. Since neuroimaging findings alone are not sufficient to support FPl’s role in complex choice, in this study we employed transcranial magnetic stimulation (TMS) to further provide causal evidence. TMS was applied to temporally disrupt the FPl or the control site vertex of human participants, followed by a two-stage decision making task. Each block of the task involved a stage of complex choice followed by a stage of simple choice (Fig. 1). In complex choice, participants chose between two complex options – each was an aggregate of 20 gambles that were associated with different rewards. In contrast, in Stage 2 they chose between two single gambles – each consisted of a number denoting the reward magnitude and a bar of variable height denoting the reward probability. Our results supported the causal role of FPl in complex choice by showing that disruption of FPl impaired decision making in complex choice, but not in simple choice nor when TMS was applied to the control vertex area. Next, to inspect the precise computations that were disrupted, participants’ behaviour was fitted with 14 computational models. The winning model suggests that participants arbitrated between using a computationally complex process (i.e. extracting multiple features of the option, namely the mean, variance, and skewness) and a simple heuristic (i.e. concerning only the mean) for integrating choice information. Crucially, participants became less likely to use the complex process after FPl disruption. In addition, in a separate neuroimaging experiment, we found that participants with stronger FPl signals also tended to use the complex process more. Taken together, our findings provide causal evidence of FPl’s role in information integration during complex decision making.