Algorithmically-mediated content is both a product and producer of dominant social narratives, and it has the potential to impact users’ beliefs and behaviors. We present two studies on the content and impact of gender and racial representation in image search results for common occupations. In Study 1, we compare 2020 workforce gender and racial composition to that reflected in image search. We find evidence of underrepresentation on both dimensions: women are underrepresented in search at a rate of 42% women for a field with 50% women; people of color are underrepresented with 16% in search compared to an occupation with 22% people of color (the latter being proportional to the U.S. workforce). We also compare our gender representation data with that collected in 2015 by Kay et al., finding little improvement in the last half-decade. In Study 2, we study people’s impressions of occupations and sense of belonging in a given field when shown search results with different proportions of women and people of color. We find that both axes of representation as well as people’s own racial and gender identities impact their experience of image search results. We conclude by emphasizing the need for designers and auditors of algorithms to consider the disparate impacts of algorithmic content on users of marginalized identities.

The task of definition detection is important for scholarly papers, because papers often make use of technical terminology that may be unfamiliar to readers. Despite prior work on definition detection, current approaches are far from being accurate enough to use in real- world applications. In this paper, we first perform in-depth er- ror analysis of the current best performing definition detection system and discover major causes of errors. Based on this analysis, we develop a new definition detection system, HEDDEx, that utilizes syntactic features, transformer encoders, and heuristic filters, and evaluate it on a standard sentence-level benchmark. Because current benchmarks evaluate randomly sampled sentences, we propose an alternative evaluation that assesses every sentence within a document. This allows for evaluating recall in addition to precision. HEDDEx outperforms the leading system on both the sentence-level and the document-level tasks, by 12.7 F1 points and 14.4 F1 points, respectively. We note that performance on the high-recall document-level task is much lower than in the standard evaluation approach, due to the necessity of incorporation of document structure as features. We discuss remaining challenges in document-level definition detection, ideas for improvements, and potential is- sues for the development of reading aid applications.

Despite the central importance of research papers to scientific progress, they can be difficult to read. Comprehension is often stymied when the information needed to understand a passage resides somewhere else—in another section, or in another paper. In this work, we envision how interfaces can bring definitions of technical terms and symbols to readers when and where they need them most. We introduce ScholarPhi, an augmented reading interface with four novel features: (1) tooltips that surface position-sensitive definitions from elsewhere in a paper, (2) a filter over the paper that “declutters” it to reveal how the term or symbol is used across the paper, (3) automatic equation diagrams that expose multiple definitions in parallel, and (4) an automatically generated glossary of important terms and symbols. A usability study showed that the tool helps researchers of all experience levels read papers. Furthermore, researchers were eager to have ScholarPhi’s definitions available to support their everyday reading.