Do people sympathize with ostracized AI agents without blaming humans?

In the same Cyberball experiment (N=244) reported in Do people treat excluded AI agents like excluded humans?, Zhou, Porat and Van Zalk measured a second, separate response alongside the behavioral compensation finding: self-reported sympathy toward the coplayers and evaluations of the human player's conduct. The authors report that "participants showed increased sympathy towards the ostracised agent, but they did not devalue the human player for their ostracising behaviour; this indicates that participants did not mindfully perceive AI virtual agents as comparable to humans."

The authors read this pairing as evidence that the mindless and mindful routes identified by the dual-processing model (Type 1 automatic, Type 2 reflective) can point in different directions within the same observer and the same episode. The behavioral measure (ball-tossing frequency) showed participants mindlessly treating the agent like a human worth including; the evaluative measure (judging the human ostracizer) showed that when participants consciously, reflectively assessed the human player's behavior, they did not extend the agent the same moral standing — ostracizing an agent was not treated as comparable to ostracizing a person, so the human player escaped the social penalty a human-human Cyberball study would predict. Sympathy rose, but social blame did not transfer.

This qualifies the central claim rather than contradicting it: the paper's title captures the mindless-norm result, but this second finding is what licenses the paper's own conclusion that the "social being" perspective is only partially supported. It also bears on Do humans apply human-human scripts to AI interactions? — if people maintain a reflective category boundary between agents and humans even while mindlessly applying human-human behavioral scripts to agents, that supports the idea of two coexisting script or processing systems rather than one unified treat agents as humans response. It is also a caution for the mindless/mindful anthropomorphism distinction used in Do more social cues always make AI feel more present?: both forms of anthropomorphism can be present and diverge in direction within a single interaction, not just independently activate.

The study used self-report sympathy and evaluation measures collected after a single staged observation of ostracism in a ball-tossing game; it does not establish whether the sympathy-without-blame pattern holds for direct (not third-person) human-agent interaction, for agents with richer social cues than Cyberball's minimal coplayer role, or outside the specific moral frame of "ostracism." The paper also reports two further exploratory findings not detailed in this excerpt — an association between frequency of agent usage and sympathy, and a carryover effect of positive usage experience — that the authors flag as exploratory and that this note does not take as established.