Abstract Human–bot collaborative knowledge building is increasingly prevalent in online knowledge communities (OKCs). However, it is unclear how human and bot individual activities and their collaboration influence different stages of collective knowledge evolution. Based on a dataset of 337,830 edit and talk records sampled from 200 randomly selected Wikipedia articles, we first conducted an exploratory analysis to examine how the frequency of human and bot activities and the “frequency–breadth–depth” of their collaboration patterns change across stages of collective knowledge evolution. Furthermore, we utilized a fixed‐effect multinomial logistic model to assess the effects of these variables on the stages of collective knowledge evolution. The results indicate the distinct impacts of human/bot activities and human–bot collaboration patterns on idea generating, idea organizing, and intellectual convergence. The core findings include (1) human activities promote idea generating and organizing, whereas bot activities facilitate idea organizing; (2) human‐led collaboration patterns enhance idea generating and organizing, while co‐led collaboration patterns promote intellectual convergence. These findings contribute to information science and human–bot collaboration research, as well as guide platform design and community administration of OKCs. Specifically, we recommend the implementation of stagedependent bot deployment strategies and human–bot collaboration incentives that align with the stages of collective knowledge evolution.
Abstract This paper examines the relationship between the genderedness of authors' first names and citation distributions in US‐affiliated scholarly production from 2010 to 2019. Merging a first name genderedness table derived from Wikidata with Web of Science publication data, we develop a relative distributional framework that compares first name, article, and citation distributions by both disciplinary group and author role along a continuous genderedness spectrum. The lexical structure of the corpus proves stable across author roles, while productivity diverges substantially by disciplinary group: physical sciences show a consistent masculine lean, social sciences a feminine one, with amplitude remaining stable across groups despite these directional differences. Once productivity is controlled for, citation analyses net of publication volume reveal a pervasive deficit across disciplinary groups and author roles: only unambiguously masculine names accumulate citations in excess of their publication volume, while feminine, neutral, and moderately masculine names all fall short of it. This asymmetry is strongest in the life sciences and weakest in the physical sciences except among middle authors, where the pattern reflects the gender composition of large collaborative teams rather than evaluative bias in citing behavior. Taken together, these results are consistent with the hypothesis that first name genderedness shapes citation recognition through implicit bias operating in low‐deliberation evaluative contexts.
Abstract Geographical representation of reviewers is a critical dimension of disparity in the peer review process. To investigate its relationship with gatekeeping behavior, this paper uses manuscript metadata from Elsevier's peer review platform, supplemented by national‐level indicators of economic development and of political and cultural dimensions. Accounting for national differences in publication output, we develop country‐level metrics of peer review representativeness and review duration. Our analysis reveals that, compared to North America and Western Europe, countries in Asia and Africa remain markedly underrepresented in gatekeeping. We also observe that reviewers based in Central America and Africa tend to exhibit slower review turnaround times. To further elucidate the structural underpinnings behind these disparities, we employ regression analyses linking national characteristics to gatekeeping behavior. Results suggest that countries that are overrepresented in gatekeeping tend to share similar political and cultural profiles. Specifically, countries with stronger governance frameworks tend to be structurally overrepresented as gatekeepers; editorial processes appear faster in cultures with a stronger orientation toward planning, whereas reviewers from countries characterized by higher levels of risk preference tend to show lower sensitivity to time constraints. These findings provide new empirical support for calls to advance equity in global academic participation.
Abstract Scholarly blogs are an important venue for scholarly communication; however, they are not yet integrated into the preservation workflows of digital research and information infrastructures, which places their long‐term access at risk. By using Star and Ruhleder's (1996) dimensions of information infrastructure as a theoretical framework, in this study we investigate social and organizational factors that impact blog preservation and derive implications for the implementation of blogs into an information infrastructure. We conducted and qualitatively analyzed 13 semi‐structured interviews with scholarly bloggers to identify bloggers' perceived challenges concerning blog preservation and their requirements for an information infrastructure tasked with preserving scholarly blogs. We found that participants named or indirectly described existing infrastructure institutions and their services that are already working toward preserving the scholarly record. The findings suggest extending the scope of existing infrastructure institutions to include scholarly blog preservation.
Abstract Scholarship on documentation of datasets is increasingly concerned with capturing the context, socio‐technical processes, and decisions shaping data. “Datasheets for datasets” is a recent approach developed by researchers in machine learning that foregrounds transparency and uncovering bias through data documentation. Recognizing that machine learning documentation is already being applied for cultural heritage collections as data, we assess the effectiveness of applying datasheets to document web archives datasets. Our study involved community workshops to determine priorities for what to include in a datasheet. We then developed three sample datasheets for datasets from the UK Web Archive, observing and recording the process. Our findings identify the resources needed to create a datasheet, outlining the effort required in terms of time, roles and expertise required. Findings also reveal several notable tensions that emerged in the process, locating mismatches between a framework from machine learning and applications in cultural heritage settings. We develop five provocations to prompt critical consideration of assumptions about how datasets are created and used that might be embedded in any documentation framework. We conclude with an agenda for future research engaging with data documentation in web archives, cultural heritage collections, and data curation.
