Juan E. Gilbert’s research while affiliated with University of Florida and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (238)


‘Smart’ Choice? Evaluating AI-Based mobile decision bots for in-store decision-making
  • Article

October 2024

·

21 Reads

·

2 Citations

Journal of Business Research

·

·

·

[...]

·

Juan E. Gilbert

Publisher Correction: Increasing the presence of BIPOC researchers in computational science
  • Article
  • Publisher preview available

October 2024

·

1 Read

Nature Computational Science

View access options

Two-Step Ballot Verification: Mitigating the Impact of the Hawthorne Effect on Vote-Flipping Studies

September 2024

·

5 Reads

Proceedings of the Human Factors and Ergonomics Society Annual Meeting

This paper investigates the effects of the Hawthorne effect on voter behavior in ballot verification studies, specifically in the context of using ballot marking devices (BMDs) to print paper ballots. Previous studies suggested an insufficient number of voters verify the printed ballot from the BMD. This study introduces a two-step verification process aimed at reducing the Hawthorne effect’s impact, which suggests that individuals alter their behavior due to the awareness of being observed. The methodology involves direct questioning about participants’ awareness of manipulated votes and a subsequent inquiry to identify the specific contest where a vote was flipped. The findings indicate that when directly asked, a higher percentage of participants acknowledged noticing vote discrepancies, illustrating the potential influence of the Hawthorne effect in previous research methodologies. The paper recommends a vote flipping study protocol to account for the effect and to ensure the accuracy of vote flipping studies.






Fig. 2 The outcomes by ethnicity from 10 runs of 5-fold cross-validation for multiple machine learning models and feature testing methods. a The false positive rate of each model by ethnicity as an average of the 50 runs. b The false negative rate of each model by ethnicity as an average of the 50 runs. Outliers are excluded. Values for boxplots are available in Supplementary Tables 3 and 4.
Fig. 3 Ethnicity-specific training of the best overall model (SVM). Data was collected using 10 runs of 5-fold cross-validation. a Precisionrecall curves for each model trained. b, d The balanced accuracy (b) and average precision (d) of each training as an average of the 50 runs. Bold means the highest performance among methods (columns), and underline means the highest performance among ethnicities (rows). c, e Boxplots showing the median, upper quartile and lower quartile of the balanced accuracy (c) and average precision (e). Outliers are excluded. Values for boxplots are available in Supplementary Tables 4 and 5.
Fig. 6 A heatmap showing the top ~50 features selected by the T Test method for the SVM model. The intensity of the heatmap shows higher significance. The corresponding p-values are log transformed. Features are ordered in alphabetical order. Values for the heatmap are available in Supplementary Table 12.
Ethnic disparity in diagnosing asymptomatic bacterial vaginosis using machine learning

November 2023

·

33 Reads

·

5 Citations

npj Digital Medicine

While machine learning (ML) has shown great promise in medical diagnostics, a major challenge is that ML models do not always perform equally well among ethnic groups. This is alarming for women’s health, as there are already existing health disparities that vary by ethnicity. Bacterial Vaginosis (BV) is a common vaginal syndrome among women of reproductive age and has clear diagnostic differences among ethnic groups. Here, we investigate the ability of four ML algorithms to diagnose BV. We determine the fairness in the prediction of asymptomatic BV using 16S rRNA sequencing data from Asian, Black, Hispanic, and white women. General purpose ML model performances vary based on ethnicity. When evaluating the metric of false positive or false negative rate, we find that models perform least effectively for Hispanic and Asian women. Models generally have the highest performance for white women and the lowest for Asian women. These findings demonstrate a need for improved methodologies to increase model fairness for predicting BV.




Citations (59)


... Second, Smarr & Gilbert[60] conducted a systematic literature review examining diversity initiatives geared towards Black students in undergraduate computing contexts, with a particular emphasis on the theoretical frameworks used (or not) to inform these initiatives. In line with the present study, Smarr & Gilbert specifically included studies focused on Black students in undergraduate computing, and excluded those concerned with STEM education in general. ...

Reference:

Bolstering the Persistence of Black Students in Undergraduate Computer Science Programs: A Systematic Mapping Study
Higher Education Computing Curriculum for the Black Community: A Review
  • Citing Conference Paper
  • August 2022

... Plenty of studies have addressed the existence of unfairness across multiple imaging modalities (MRI 3,4 , X-ray 2,5,6 ) and body parts (brain [7][8][9] , chest 10 , heart 3,4 , skin 11,12 ), and different sensitive attributes (sex 9,13 , age 14,15 , race 16,17 , skin tone 18,19 ). Besides, this issue is also found in other healthcare applications where the inputs to the system are electronic medical records 20,21 or RNA sequences 22 . ...

Ethnic disparity in diagnosing asymptomatic bacterial vaginosis using machine learning

npj Digital Medicine

... These models compete against each other, with the adversarial model challenging the classifier to improve its fairness, resulting in enhanced overall performance (e.g., [41,92,119]). Additionally, others leveraged unlabeled data to learn fair representations, combining techniques like resampling and reweighting (e.g., [15,58,60,89,91]). ...

Add-Remove-or-Relabel: Practitioner-Friendly Bias Mitigation via Influential Fairness
  • Citing Conference Paper
  • June 2023

... There are several critical frameworks used to define and understand disability, including autism (Lawson and Beckett, 2021). While these frameworks are ever-evolving, the medical model framework which defines disability as a "disease" is one of the most commonly applied in computer science research when discussing disability in general and autism specifically (Rizvi et al., 2024;Spiel et al., 2019a;Sideraki and Drigas, 2021;Anagnostopoulou et al., 2020;Parsons et al., 2020;Williams et al., 2023;Sum et al., 2022). This framework defines autism as a deficit of skills such as the Theory of Mind, and its applications in technology research largely focus on providing diagnosis and treatment to autistic people (Baron-Cohen, 1997;Begum et al., 2016;Rizvi et al., 2024;Spiel et al., 2019b). ...

Counterventions: a reparative reflection on interventionist HCI
  • Citing Conference Paper
  • April 2023

... A list of Black and Latinx computing doctorate recipients (n = 144) were obtained using prior research on Black faculty in CS research departments [45] and information from the Hispanics in Computing website [46]. Individuals from [45], [46] were solely identified as Black or Latinx computing scholars, so this research retained those labels in the analyses that follow and cannot ascertain if there were any Afro-Latinos in the sample. ...

An Analysis of Black Faculty in CS Research Departments
  • Citing Article
  • January 2023

Communications of the ACM

... To attain this objective, several BCI applications have been presented in the literature. including character spelling [4,5], word typing [6], controlling a wheelchair [7,8], controlling a robotic/prosthetic limb [9], virtual reality control [10,11], neurorehabilitation [12], controlling a car [13], web browser control [14], Unmanned Aerial Vehicle (UAV) control [15,16], and games [17][18][19]. All these BCIs take commands from the brain and transform them to control signals for the desired application. ...

Mind Games: A Web-Based Multiplayer Brain-Computer Interface Game
  • Citing Article
  • October 2022

Proceedings of the Human Factors and Ergonomics Society Annual Meeting

... In this context, even autonomous vehicles (AVs) might improve rural accessibility in the following decades (e.g. Bernhart et al., 2018;Dianin et al., 2021;Imhof et al., 2020;Prioleau et al., 2021). For instance, standard public transport (PT) could profit from driver cost savings and become more frequent and widespread (Mouratidis and Cobeña Serrano, 2021), while on-demand services like shared taxis could become the standard for the most remote and dispersed settlements (von Mörner, 2019). ...

Autonomous Vehicles in Rural Communities: Is It Feasible?
  • Citing Article
  • September 2021

IEEE Technology and Society Magazine

... In case the sentiment and emotion analysis relies upon training data, the data might originate from other domains, may include biases and stereotypes (66,67). Biased or unrepresentative data can lead to inaccurate sentiment analysis and inappropriate responses. ...

Examining Bias in Sentiment Analysis Algorithms Interacting with Emojis with Skin Tone Modifiers
  • Citing Chapter
  • June 2022

Communications in Computer and Information Science

... • Conducting multiple iterations of the process to achieve saturation (Dinh et al., 2023;Faber et al., 2022;Hegde et al., 2023;Hirzle et al., 2023;Impedovo et al., 2013;Turakhia et al., 2023) • Utilising multiple persons in the process of developing and/or evaluating codes (Çolakoglu et al., 2023;Dinh et al., 2023;Faber et al., 2022;Hegde et al., 2023;Hirzle et al., 2023;Impedovo et al., 2013;Rajaram et al., 2023;Turakhia et al., 2023) A review of 76 papers published between 2019 and 2023 identified in a PRISMA study on XR development strategies and policies did not reveal any codes for developing an XR project. Within these data, the researchers developed codes for the recordings of participants' responses (Williams, 2020), literature on AR use cases, benefits, or obstacles (Nassereddine, 2019), and interview transcripts (Karre et al., 2019). This finding does not prove the nonexistence of specific codes for an XR development strategy given the scope of the PRISMA study. ...

MetaCogs: Mitigating Executive Dysfunction via Agent-Based Modeling for Metacognitive Strategy Development
  • Citing Article
  • April 2022

ACM Transactions on Accessible Computing

... RTM differs from place-based forecasting in that it can identify the environmental risk factors associated with a type of crime and assess their spatial influence on an area's vulnerability to experiencing that type of crime in the future (Davis et al. 2022). Additionally, RTM does not require the inclusion of prior criminal incidents to identify areas spatially vulnerable to crime, which may temper the concern that police officer bias drives crime forecasts. ...

Five ethical challenges facing data-driven policing

AI and Ethics