Thanks for the kind words!

Led by Xiaohuan Xia in a fantastic (and super fun!) collaborative team also including @academicmatou.bsky.social, Shubhankar Patankar, and @dianatamir.bsky.social!

Looking to the future, important open questions remain. How might citation sentiment have changed over time? How might the relation between sentiment and culture be impacted by serendipity or paradigm shifts? How might other aspects of affect manifest in science communication more broadly?

Uncovering the social construction of science is not only a scholarly contribution but also an ethical one, as obscuring that construction is an epistemic harm. onlinelibrary.wiley.com/doi/10.1111/...

Here we uncover human subjectivities in scientific citation, demonstrating that citation sentiment in neuroscience tracks multiscale sociocultural norms of status, collaboration, discipline, and country.

Our study underscores the humanity that scholars bring to the scientific enterprise. Individually, we bring our epistemological standpoints, cognitive biases, perceptions, & values. In groups, we bring disciplinary norms of preferred practices, methods, standards, & explanations.

We find that men cite with greater sentiment---both critical & favorable---& women cite with greater sentiment bias, being more favorable to collaborators than to non-collaborators. We discuss these differences in light of prior evidence uncovering gender differences in social engagement in science.

Yet, an element that can influence all of these factors is personal identity: dimensions of identity such as gender, sex, race, ethnicity, class, and (dis)ability can determine one's placement, status, and power in social groups across scales. How might citation sentiment track identity?

Thus far, the data consistently suggest that citation sentiment can track ingroup/outgroup relations, structures of dominance, & hierarchies that are interpersonal (collaboration), prestigious (h-index, disciplinary level of explanation), and national (beliefs about power and society).

We find that citation sentiment tracks the degree to which countries accept the unequal distribution of power. Results suggest that norms of social hierarchy may influence citation sentiment, resulting in stronger criticism in high-individualism cultures that reject the uneven distribution of power.

So far, we considered a scholar's location within social structures from small collaborations to wider disciplines. But the sociocultural milieu in which each scholar exists also extends outwards to the national scale, with each country having distinct norms that can manifest in scientific practice.

Moving from dyads to groups, we consider citation sentiment across scientific disciplines. We find that drylab disciplines use more sentiment (favorable & critical), whereas wetlab disciplines cite more neutrally. These findings track the differently valued explanations that disciplines offer.

How might citation sentiment track structures of dominance? We consider the hierarchy reflected in the h-index. We find that greatest critical sentiment appears when higher h-index scholars cite lower h-index scholars, demonstrating that sentiment varies with relative status of the citer vs. citee.

How might sentiment align with ingroup/outgroup relations? We considered the ingroup/outgroup relations of collaborators vs. non-collaborators. We find that we cite ourselves most favorably, collaborators next favorably, and non-collaborators most critically.

To ensure that our assessment is computationally tractable while remaining sensitive to local sociocultural norms, we focus on a single science---neuroscience---and examine citation sentiment in >100K articles from 181 journals, 27 departments, and 23 countries and regions.

We ask not about who we cite but how we cite. Using a large language model, we evaluate citation sentiment (favorable, critical, or neutral) and how it tracks social norms of collaboration, discipline, and culture. Our study provides a paper-trail lens into the socio-cognitive processes of science.