In human society, sharing losses is at least as common as sharing gains. Although the psychological and neural processes underlying the latter have been investigated in depth, those related to the former are not clear. Our recent study demonstrates an increased demand for fairness under adversity (e.g. loss sharing). Here we investigated how our brain encodes unfairness in the loss and gain domains using event-related potentials (ERP) technique. We adopted the Ultimatum Game (UG) to probe the processes related to fairness consideration in either gain or loss domain. In UG, two players, the proposer and the responder, bargain on how to divide a certain amount of money endowed by the experimenter. The proposer suggests a division policy, on which the responder evaluates and decides whether to accept. Upon acceptance, the money is divided as suggested; while rejection results in both players going empty-handed. Participants, as responders, were required to decide whether to accept an offer that was either fair (equal or nearly equal division) or unfair in both gain and loss domain. Offers were either made by the human partner or by the computer partner. Behavioral results replicated our previous findings that the rejection rate of unfair offers was higher in the loss than in the gain domain. ERP results revealed that the N1 amplitude was more pronounced for human partners compared with computer partners, however, this effect was only observed in the gain domain. When interacting with computer partners, unfair offers and offers in the loss domain were associated with larger N350 compared with fair offers and offers in the gain domain, and offers in the gain domain elicited larger P2 than offers in the loss domain, whereas fair offers and offers in the gain domain were associated with larger LPP than unfair offers and offers in the loss domain. In addition, these differences in ERP responses were diminished when the interacting partners were humans. These findings suggest that fairness processing is modulated by the property of the partner and gain-loss domain. In human-computer interaction, unfair offers and offers in the loss domain elicit more inhibition and conflict resolving process, while fair offers and offers in the gain domain are more motivationally significant to human. The present findings support the view that fairness processing is context-dependent, in which factors like gain-loss domain and the property of the partner play a role.