Epistemic Reciprocity in Emergencies: Why the way we reciprocate information matters

1. Shared identity in emergencies

In real-world crises, how people exchange and reciprocate information can be critical for saving lives. Information transmission is particularly at high stakes as it plays a vital role in effective cooperations among multiple parties, including first responders and, zero responders (survivors at the scene), to help the victims (e.g., injured individuals). Given the increasing use of AI-enabled systems to support emergency response, we argue that patterns of information exchange and reciprocity can serve as key indicators of the psychological state of groups on the ground. In turn, these indicators can inform system decision-making, thereby supporting more effective emergency response and collaboration.

Epistemic reciprocity can enhance the accuracy of shared identity detection architectures, improving system responsiveness to group behaviours at the service of saving lives.

Social psychological research has shown that, during emergencies, people affected by the emergency tend to psychologically shift from “me in relation to other individuals” to “us in relation to the emergency.” This enhances perceptions of similarity and bonding due to the “common fate” and becomes the basis of an emergent shared identity — a sense of groupness among survivors (Drury, 2018). Survivors acting based on a shared identity, tend to exhibit more prosocial behaviours (helping, providing mutual support, and coordinating action). Applied to the design of an emergency response AI system, such a shared identity can be detected in verbal and non-verbal behaviour and adapt the system’s decision-making. Based on simulations (Gavidia-Calderon et al., 2024) and synthetic data (Kordoni et al., 2023), research shows that identity adaptive autonomous systems (e.g., drones) can achieve larger number of evacuations than identity-non adaptive systems.

Our empirical research discovered that epistemic reciprocity – a concrete linguistic metric of information transmission and reciprocity — could indicate psychological groupness, i.e., shared identity. A concrete linguistic metric —epistemic reciprocity can enhance the accuracy of shared identity detection architectures, improving system responsiveness to group behaviours at the service of saving lives.

2. Linguistic epistemic reciprocity as an indicator of shared identity

Epistemic reciprocity measures how information is transmitted and reciprocated among information givers and information receivers. As the social identity approach posits that people with shared social identity tend to align their behaviours with the group norms, their speech acts of information transmission should also align with the group norms. 
To ensure the robustness of the linguistic metric, we studied high quality recordings of real-life interactions in emergencies from Emergency Helicopter Medics docuseries. 868 speech acts of information transmission in total were transcribed and annotated following Tantucci et al.'s (2022) work on epistemic reciprocity, incorporating social identity theory.

There are three levels of epistemic reciprocity, which are downward shift (D), matching (M) and upward shift (U) of epistemic reciprocity, according to the linguistic structures and forms of interactions. Individuals who are performing the information transmission are annotated as information receiver and information giver. Their identity as either first responder or zero responder, and their institutional status are also annotated. Once all data is annotated, we then employed Multiple Correspondence Analysis (MCA). 

... people within a shared group membership and people from different group memberships reciprocate information differently during emergencies.

The results indicated that people within a shared group membership and people from different group memberships reciprocate information differently during emergencies.
We found two clusters associating with different identities. The first cluster demonstrated an emergent shared identity between first responders and zero responders, where people reciprocate information transmission with more effort (upward shift in epistemic reciprocity). The second cluster showed an alignment of a shared psychological state between first responders who cooperate in the same emergencies, where first responders maintain the status quo when reciprocating information transmission (matching level of epistemic reciprocity). 

3. Implications

Our results have demonstrated that a shared identity in emergencies can be manifested linguistically in epistemic reciprocity: an upward shift of epistemic reciprocity indicates the formation of emergent shared identity, whereas matching epistemic reciprocity indicates the maintenance of an established shared identity. It contributes to a psychologically informed coordination by detecting when collaborative momentum is forming or disrupted, and building capacity to adapt operational interventions accordingly. The development and deployment of responsible AI-enabled systems using such epistemic reciprocity notions can better align with collective identity signals, enabling them to make decisions based on group-level dynamics rather than isolated individual features during emergencies.