William H. Walters

Dr. William H. Walters

orcid

Research Focus & Metrics

Academic Career Length

28Years (1999-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
7 September 2023 in Scientific Reports
Verified
Journal Article
Open access
PDF file
WWWilliam H. Walters
William H. Walters
EWEsther Isabelle Wilder
Esther Isabelle Wilder
Although chatbots such as ChatGPT can facilitate cost-effective text generation and editing, factually incorrect responses (hallucinations) limit their utility. This study evaluates one particular type of hallucination: fabricated bibliographic citations that do not represent actual scholarly works. We used ChatGPT-3.5 and ChatGPT-4 to produce short literature reviews on 42 multidisciplinary topics, compiling data on the 636 bibliographic citations (references) found in the 84 papers. We then searched multiple databases and websites to determine the prevalence of fabricated citations, to identify errors in the citations to non-fabricated papers, and to evaluate adherence to APA citation format. Within this set of documents, 55% of the GPT-3.5 citations but just 18% of the GPT-4 citations are fabricated. Likewise, 43% of the real (non-fabricated) GPT-3.5 citations but just 24% of the real GPT-4 citations include substantive citation errors. Although GPT-4 is a major improvement over GPT-3.5, problems remain.

1 January 2023 in Open Information Science
Verified
Journal Article
Open access
PDF file
WWWilliam H. Walters
William H. Walters
This study evaluates the accuracy of 16 publicly available AI text detectors in discriminating between AI-generated and human-generated writing. The evaluated documents include 42 undergraduate essays generated by ChatGPT-3.5, 42 generated by ChatGPT-4, and 42 written by students in a first-year composition course without the use of AI. Each detector’s performance was assessed with regard to its overall accuracy, its accuracy with each type of document, its decisiveness (the relative number of uncertain responses), the number of false positives (human-generated papers designated as AI by the detector), and the number of false negatives (AI-generated papers designated as human). Three detectors – Copyleaks, TurnItIn, and Originality.ai – have high accuracy with all three sets of documents. Although most of the other 13 detectors can distinguish between GPT-3.5 papers and human-generated papers with reasonably high accuracy, they are generally ineffective at distinguishing between GPT-4 papers and those written by undergraduate students. Overall, the detectors that require registration and payment are only slightly more accurate than the others.

21 April 2021 in Scholarly Assessment Reports
Verified
Journal Article
Open access
PDF file
EWEsther Isabelle Wilder
Esther Isabelle Wilder
WWWilliam H. Walters
William H. Walters
Although large citation databases such as Web of Science and Scopus are widely used in bibliometric research, they have several disadvantages, including limited availability, poor coverage of books and conference proceedings, and inadequate mechanisms for distinguishing among authors. We discuss these issues, then examine the comparative advantages and disadvantages of other bibliographic databases, with emphasis on (a) discipline-centered article databases such as EconLit, MEDLINE, PsycINFO, and SocINDEX, and (b) book databases such as Amazon.com, Books in Print, Google Books, and OCLC WorldCat. Finally, we document the methods used to compile a freely available data set that includes five-year publication counts from SocINDEX and Amazon along with a range of individual and institutional characteristics for 2,132 faculty in 426 U.S. departments of sociology. Although our methods are time-consuming, they can be readily adopted in other subject areas by investigators without access to Web of Science or Scopus (i.e., by faculty at institutions other than the top research universities). Data sets that combine bibliographic, individual, and institutional information may be especially useful for bibliometric studies grounded in disciplines such as labor economics and the sociology of professions. Policy highlightsWhile nearly all research universities provide access to Web of Science or Scopus, these databases are available at only a small minority of undergraduate colleges. Systematic restrictions on access may result in systematic biases in the literature of scholarly communication and assessment.The limitations of the largest citation databases influence the kinds of research that can be most readily pursued. In particular, research problems that use exclusively bibliometric data may be preferred over those that draw on a wider range of information sources.Because books, conference papers, and other research outputs remain important in many fields of study, journal databases cover just one component of scholarly accomplishment. Likewise, data on publications and citation impact cannot fully account for the influence of scholarly work on teaching, practice, and public knowledge.The automation of data compilation processes removes opportunities for investigators to gain first-hand, in-depth understanding of the patterns and relationships among variables. In contrast, manual processes may stimulate the kind of associative thinking that can lead to new insights and perspectives.

9 October 2017 in IEEE Access
Verified
Journal Article
Open access
WWWilliam H. Walters
William H. Walters
This guide presents nine key questions that can help researchers make good use of citation-based journal rankings (metrics) in the natural and social sciences. The nine questions address the characteristics that distinguish one metric from another: the source documents, the citation-counting window, the document types counted, the cited-document window, the impact of highly cited documents, the treatment of self-citations, the distinction between size-dependent and size-independent metrics, the use of normalization to account for disciplinary differences in impact, and the use of weighting to account for the impact or centrality of each citing journal. Next, the guide reviews 19 standard citation metrics, including the $h$ index, $g$ index, impact factor, source normalized impact per paper, eigenfactor, article influence score, and SCImago journal rank. Three underlying data sources (Web of Science, Scopus, and Google Scholar) are described, along with six major data download sites: Journal Citation Reports, Eigenfactor, CWTS Journal Indicators, SCImago, Scopus Journal Metrics, Cabell’s International, and Google Scholar Metrics. The paper summarizes the main criticisms of citation metrics and concludes with suggestions for their further development, dissemination, and use.


29 April 2015 in Journal of the Association for Information Science and Technology
Verified
Journal Article
Closed access
PDF file
WWWilliam H. Walters
William H. Walters
EWEsther Isabelle Wilder
Esther Isabelle Wilder