Prosocial traffic behavior – yielding, coordinating, and signaling to reduce others’ risk – underpins human–automation co-existence in mixed traffic. However, real-world evidence on when and toward whom automated vehicles (AVs) should act prosocially is scarce. A mixed-methods field study in San Francisco combined 13 naturalistic observations (11 in-ride, 2 curbside) with 28 semi-structured interviews with pedestrians, cyclists, and drivers conducted within SALSA, a project funded by the German Federal Ministry for Economic Affairs and Energy. The analysis centered on prosocial AV-interactions, synthesizing expectations, experiences, and observed interaction patterns.

Expectations were strongly role dependent: 77% of pedestrians and all cyclists prioritized defensive, prosocial AV behavior, whereas 42% of drivers did so, emphasizing predictability and assertiveness to sustain traffic flow. Observations mirrored this asymmetry: AVs were consistently defensive with vulnerable road users but displayed more assertive maneuvers in vehicle-to-vehicle interactions, three judged safety-critical (e.g., tight passes at bottlenecks). Drivers reported exploiting overly cautious AVs (e.g., cutting in), revealing a reciprocity gap that penalizes one-sided prosociality.

This role-sensitive adaptation may help explain positive evaluations of AVs in San Francisco. However, lacking communication was a recurring concern, detracting from perceived prosociality. eHMI cues were noticed by only a minority of interviewees, undermining coordination in close-range encounters. The study supports a context-sensitive framework of prosocial assertiveness: (a) deferential behavior toward vulnerable users; (b) calibrated, norm-consistent assertiveness with other vehicles to prevent exploitation; and (c) eHMI that clearly communicates recognition and intent. These insights yield concrete behavioral and interface targets to foster cooperative, safe, legible mixed traffic.