Unpacking the black box

Martins, Thiago Gonçalves dos Santos; Schor, Paulo

doi:10.31744/einstein_journal/2021ed6037

Cited by 3 publications

(2 citation statements)

References 9 publications

(9 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Future analyses should compare metrics between preprinted articles and a comparator group of similar articles published simultaneously that are alike in terms of orthopaedic topic and study design but are not the same study as the preprinted article. Additionally, previous analyses have shown that the Altmetric score is mainly driven by Twitter [4, 5, 15], suggesting this analysis counts tweets twice through these two metrics. However, Altmetric helps provide a different angle of the Twitter amplification metric, specifically through its ability to disambiguate links to different research outputs in tweets [2].…”

Section: Discussionmentioning

confidence: 94%

Characterization and Reach of Orthopaedic Research Posted to Preprint Servers: Are We “Undercooking” Our Science?

Rizk

Arza

Jella

et al. 2023

Clin Orthop Relat Res

View full text Add to dashboard Cite

Background Although biomedical preprint servers have grown rapidly over the past several years, the harm to patient health and safety remains a major concern among several scientific communities. Despite previous studies examining the role of preprints during the Coronavirus-19 pandemic, there is limited information characterizing their impact on scientific communication in orthopaedic surgery. Questions/purposes (1) What are the characteristics (subspecialty, study design, geographic origin, and proportion of publications) of orthopaedic articles on three preprint servers? (2) What are the citation counts, abstract views, tweets, and Altmetric score per preprinted article and per corresponding publication? Methods Three of the largest preprint servers (medRxiv, bioRxiv, and Research Square) with a focus on biomedical topics were queried for all preprinted articles published between July 26, 2014, and September 1, 2021, using the following search terms: “orthopaedic,” “orthopedic,” “bone,” “cartilage,” “ligament,” “tendon,” “fracture,” “dislocation,” “hand,” “wrist,” “elbow,” “shoulder,” “spine,” “spinal,” “hip,” “knee,” “ankle,” and “foot.” Full-text articles in English related to orthopaedic surgery were included, while nonclinical studies, animal studies, duplicate studies, editorials, abstracts from conferences, and commentaries were excluded. A total of 1471 unique preprints were included and further characterized in terms of the orthopaedic subspecialty, study design, date posted, and geographic factors. Citation counts, abstract views, tweets, and Altmetric scores were collected for each preprinted article and the corresponding publication of that preprint in an accepting journal. We ascertained whether a preprinted article was published by searching title keywords and the corresponding author in three peer-reviewed article databases (PubMed, Google Scholar, and Dimensions) and confirming that the study design and research question matched. Results The number of orthopaedic preprints increased from four in 2017 to 838 in 2020. The most common orthopaedic subspecialties represented were spine, knee, and hip. From 2017 to 2020, the cumulative counts of preprinted article citations, abstract views, and Altmetric scores increased. A corresponding publication was identified in 52% (762 of 1471) of preprints. As would be expected, because preprinting is a form of redundant publication, published articles that are also preprinted saw greater abstract views, citations, and Altmetric scores on a per-article basis. Conclusion Although preprints remain an extremely small proportion of all orthopaedic research, our findings suggest that nonpeer-reviewed, preprinted orthopaedic articles are being increasingly disseminated. These preprinted articles have a smaller academic and public footprint than their published counterparts, but they still reach a substantial audience through infrequent and superficial online interactions, which are far from equivalent to the engagement facilitated by peer review. Furthermore, the sequence of preprint posting and journal submission, acceptance, and publication is unclear based on the information available on these preprint servers. Thus, it is difficult to determine whether the metrics of preprinted articles are attributable to preprinting, and studies such as the present analysis will tend to overestimate the apparent impact of preprinting. Despite the potential for preprint servers to function as a venue for thoughtful feedback on research ideas, the available metrics data for these preprinted articles do not demonstrate the meaningful engagement that is achieved by peer review in terms of the frequency or depth of audience feedback. Clinical Relevance Our findings highlight the need for safeguards to regulate research dissemination through preprint media, which has never been shown to benefit patients and should not be considered as evidence by clinicians. Clinician-scientists and researchers have the most important responsibility of protecting patients from the harm of potentially inaccurate biomedical science and therefore must prioritize patient needs first by uncovering scientific truths through the evidence-based processes of peer review, not preprinting. We recommend all journals publishing clinical research adopt the same policy as Clinical Orthopaedics and Related Research ®, The Bone & Joint Journal, The Journal of Bone and Joint Surgery, and the Journal of Orthopaedic Research, removing any papers posted to preprint servers from consideration.

show abstract

Section: Discussionmentioning

confidence: 94%

Characterization and Reach of Orthopaedic Research Posted to Preprint Servers: Are We “Undercooking” Our Science?

Rizk

Arza

Jella

et al. 2023

Clin Orthop Relat Res

View full text Add to dashboard Cite

show abstract

“…In order to overcome this drawback, a ML-supplemented approach can be integrated into the simulation framework, by exploiting the high predictive power of the AI techniques together with the high interpretability and readability of simulation approaches [9][10][11][12].…”

Section: Introductionmentioning

confidence: 99%

Combining simulation models and machine learning in healthcare management: strategies and applications

Ponsiglione,

Zaffino,

Ricciardi

et al. 2024

Prog. Biomed. Eng.

View full text Add to dashboard Cite

Simulation models and artificial intelligence are largely used to address healthcare and biomedical engineering problems. Both approaches showed promising results in the analysis and optimization of healthcare processes. Therefore, the combination of simulation models and artificial intelligence could provide a strategy to further boost the quality of health services. In this work, a systematic review of studies applying a hybrid simulation models and artificial intelligence approach to address healthcare management challenges was carried out. Scopus, Web of Science, and PubMed databases were screened by independent reviewers. The main strategies to combine simulation and artificial intelligence as well as the major healthcare application scenarios were identified and discussed. Moreover, tools and algorithms to implement the proposed approaches were described. Results showed that machine learning appears to be the most employed artificial intelligence strategy in combination with simulation models, which mainly rely on agent-based and discrete-event systems. The scarcity and heterogeneity of the included studies suggested that a standardized framework to implement hybrid machine learning-simulation approaches in healthcare management is yet to be defined. Future efforts should aim to use these approaches to design novel intelligent in-silico models of healthcare processes and to provide effective translation to the clinics.

show abstract

Explainable AI for Cybersecurity

Sindiramutty,

Tan,

Lau

et al. 2024

Advances in Computational Intelligence and Robotics

View full text Add to dashboard Cite

In recent years, the utilization of AI in the field of cybersecurity has become more widespread. Black-box AI models pose a significant challenge in terms of interpretability and transparency, which is one of the major drawbacks of AI-based systems. This chapter explores explainable AI (XAI) techniques as a solution to these challenges and discusses their application in cybersecurity. The chapter begins with an explanation of AI in cybersecurity, including the types of AI commonly utilized, such as DL, ML, and NLP, and their applications in cybersecurity, such as intrusion detection, malware analysis, and vulnerability assessment. The chapter then highlights the challenges with black-box AI, including difficulty identifying and resolving errors, the lack of transparency, and the inability to understand the decision-making process. The chapter then delves into XAI techniques for cybersecurity solutions, including interpretable machine-learning models, rule-based systems, and model explanation techniques.

show abstract

Unpacking the black box

Cited by 3 publications

References 9 publications

Characterization and Reach of Orthopaedic Research Posted to Preprint Servers: Are We “Undercooking” Our Science?

Characterization and Reach of Orthopaedic Research Posted to Preprint Servers: Are We “Undercooking” Our Science?

Combining simulation models and machine learning in healthcare management: strategies and applications

Explainable AI for Cybersecurity

Contact Info

Product

Resources

About