Search

Machine learning is changing how researchers interact with scholarly literature. While it has the potential to reveal exciting new connections between areas of study, popular commercial tools that provide recommendations to users based on personalized profiles developed with machine learning can negatively impact the breadth of sources they will discover. Over time, they may even limit the questions researchers are willing to explore. Find out what machine learning is and how it is being deployed to make the literature more manageable, then learn strategies to warn users about its pitfalls and advocate for the responsible use of this technology.

The white-lipped deer (Cervus albirostris) is a rare and endangered species found in the Qinghai-Tibet Plateau in China. To understand the space occupancy, activity rhythm, and sexual segregation of the white-lipped deer, 24,096 effective photos and 827 effective videos were captured using infrared cameras from February 2020 to January 2022. The ecology and behavior of the white-lipped deer in Jiacha Gorge were studied in more detail using site occupancy models, relative abundance index, and other technologies and methods. The results show that The occupancy predicted by the model exceeds or approaches 0.5. The occupancy increases with greater altitude and with larger EVI values, while the detection rate increases with altitude only during spring and decreases with EVI values only in summer. The daily activity peaks for white-lipped deer were observed from 7:00 to 11:00 and 17:00 to 22:00, with annual activity peaks occurring from April to June and from September to November. From July to the following January, white-lipped deer mostly move in mixed-sex groups, while during the remainder of the year, they predominantly associate with individuals of the same sex. Climate, vegetation coverage, food resources, and human disturbance collectively influenced the behavior and habitat utilization of white-lipped deer. The foundational research conducted on white-lipped deer over the past two years is expected to enhance the basic understanding of white-lipped deer in the Qinghai-Tibet Plateau and contribute to future protection and management decisions.

Openness in research can lead to greater reproducibility, an accelerated pace of discovery, and decreased redundancy of effort. In addition, open research ensures equitable access to knowledge and the ability for any community to assess, interrogate, and build upon prior work. In order for research to succeed, openness and reproducibility are required. In turn, this requires open infrastructure and distributed access; but few institutions can provide all of these services alone. Providing a trustworthy network for perpetual availability of research data is critical to ensuring reproducibility, transparency, and ongoing inquiry. There is increased attention on the importance of open research and data sharing, leading to a proliferation of platforms to store data, materials, etc. These platforms exist in a fragmented environment that lacks technical integration and coordination with local library expertise and services, hampering curation and long-term stewardship. For example, the open source OSF enables researchers to directly create and manage research projects and integrates with other tools researchers use (Google Drive, Dropbox, Box, etc.), but lacks the ability to archive that materialle locally at a researcher’s institution. Long-term stewardship and preservation requires multiple copies of data archived in different locations, and creating archives seamlessly would be ideal. COS and IA propose to address the preservation and stewardship challenges by providing open, cooperative infrastructure to ensure long-term access and connection to research data, and by supporting and promoting adoption of open science practices to enhance research reproducibility as well as data sharing and reuse. In this workshop, participants will gain skills to implement reproducible research practices using OSF and other tools, and then curate and preserve research artifacts on Internet Archive. We will further demonstrate the possibilities to preserve such content in local institutional repositories or other distributed networks, uses community-developed standards and protocols.

In the summer of 2021, the head of the Department of Entomology and Plant Pathology contacted the Research Data Management Librarian seeking advice on what to do with the paper records, including original data, of two professors emeriti. In cooperation with the Agriculture Librarian, the Special Collections Librarian, and the Library Archivist, the collection was assessed and the files physically transported to the library. The data can be safely preserved in their original form within Special Collections. However, the department head and librarians all agreed that digitization, which would make the data more accessible and permit machine-readability, is the ideal solution. The Agriculture Librarian and the Research Data Management Librarian will select a subset of the paper records to scope out the time, effort, and level of expertise needed to digitize the data, translate it into machine-readable formats, and add appropriate metadata. Important questions about the process include determining the amount and type of training needed for different stages of the workflow, whether there are any steps where efficiency could be improved through automated processes, and how it can be scaled up to allow for processing the entire collection. The department head requested that we share recommendations for best practices and procedures with him as he works to develop a data archiving policy for the department. Recognizing that more faculty will retire and leave cabinets full of paper data, the librarians are eager to experiment with methods that could make this data FAIR before the problem becomes unmanageable and irreplaceable work is lost.

Those who work with data have learned the importance of provenance, documentation, standardization, context, and metadata in maintaining the quality of datasets. This was historically done to preserve their utility for human reuse and re-examination, but in recent years the emphasis on machine-readability of datasets has increased, in part to allow for their use in AI (artificial intelligence) applications. Just as those involved in creating and maintaining datasets benefit from an improved understanding of how they might be used with AI, the developers of AI systems should pay attention to issues that affect the data upon which their models rely. Several Google researchers present this perspective in “‘Everyone wants to do the model work, not the data work’: Data Cascades in High-Stakes AI”, a conference paper based on their qualitative study of AI practitioners (Sambasivan et al., 2021).

Researchers at the Auburn University Libraries are collaborating with cross-campus units and private companies to explore the application of artificial intelligence and machine learning (AI/ML) tools to scholarly repositories, structured and unstructured datasets, and the open web. The goal is twofold: to develop hands-on AI/ML expertise in the libraries and to better position the libraries to support and participate in Auburn University’s research activities. This chapter describes an exploratory project the libraries are working on with the IBM Research Triangle Park Center for Advanced Studies (IBM RTP-CAS) on using the IBM Watson Studio of AI/ML services to build natural-language query interfaces for scholarly repositories and datasets in specific subject domains. Since 2017, the libraries have been providing high-level information technology support and subject-matter expertise to the Military REACH Project, which is based in the College of Human Sciences (CHS) at Auburn University and funded by the US Department of Agriculture and the US Department of Defense. The mission of the Military REACH Project is to support US military families and family readiness by translating academic research and other resources into practical applications. A key component of the project is the design and development of a publicly accessible, easy-to-navigate library of research publications and other resources on military life and family health, delivered through a user-friendly website. With IBM’s help, the Watson suite of tools and services is being used to build an AI/ML-powered query interface and recommendation system for the Military REACH Library and public datasets in the same field. In this chapter, the authors discuss the larger library and professional context for this project, its background and rationale, the deliverables they are producing, and the technical, logistical, and administrative challenges that they have encountered in this effort. The goal is to provide other academic libraries with a model for embarking on similar projects and a clear understanding of the benefits and challenges involved.

Additional file 1: Table S1. Mean body size data of Asian common toad from sampled populations across its distribution range (ranked from north to south). Table S2. Data of environmental predictors of sampled populations. Table S3. Correlation coefficients between each environmental variable (statistically significant [P

Are institutional policies designed to advance open scholarship capable of accomplishing that end? What other consequences might they have for the practice of research? These mandates undoubtedly increase the number of publicly-available articles and datasets in the short-term, but there is also an expectation that as researchers become accustomed to the new requirements, the culture within their disciplines will shift to support open research, for instance through the founding of new open-access journals or platforms. Is this reasonable? The answer depends on the fate of individual scholars who support openness within their fields. While disciplinary norms are not biological traits, they can be said to evolve as new attitudes take root and proliferate in the population of members. Cultural change is only sustainable when those who hold the new attitudes can remain in the population. In this paper, I argue that the choices that determine professional gain and loss facing each researcher who wants to advance in their areas of study can be modeled as a prisoner’s dilemma. I first discuss why patterns in researcher behavior and beliefs suggest a prisoner’s dilemma in the area of data sharing, then apply this line of reasoning to the problem of perverse incentives in open access publishing. Finally, I explain what makes certain open access mandates counterproductive and propose alternative solutions that reward researchers who choose to make their work open. This provides a path towards reforming disciplinary norms concerning publishing.

Geophagy is common in extant nonhuman primate species, but the exact reasons for it across species remain unclear. Previous diet studies on Yunnan snub-nosed monkeys (Rhinopithecus bieti) were only focused on organic materials (plants and small animals). There are no reports on R. bieti exhibiting geophagy in the field before this study. This study was carried out at Xiangguqing in the Baimaxueshan Nature Reserve from June 2008 to May 2009. We recorded the behavior of geophagy and collected samples of soil consumed by the monkeys there and analyzed their content in a laboratory. We identified a total of eight sites where the monkeys consumed soil in the home range during the study period. The total time spent ingesting soil was 13,690 seconds. 20 adult males, 34 adult females except lactating mothers, and 12 immatures without infants were seen to eat soil throughout this study. Average time spent in soil-eating bouts differed significantly among age/sex classes. This study suggests that particular age/sex classes or individuals in certain states of society and health will predictably display a behavioral pattern of geophagy. Our data indicate that geophagy in R. bieti is best explained as a response to nutrient deficiency, as soil consumed by the monkeys was significantly higher in calcium, copper, zinc, iron, manganese, and sodium. Although R. bieti consumes iron- rich soil, control samples that were not consumed also had high levels of iron, suggesting that high altitude alone is not a sufficient explanation for geophagy in this species.