Northern Kentucky University

CoVCues: A Trustworthy Resource Amidst The COVID Infodemic

Grade Level at Time of Presentation

Senior

Major

Data Science

Minor

Applied Statistics

Institution 24-25

Northern Kentucky University

Department

Cybersecurity & Information Technology

Abstract

The COVID-19 pandemic led to an increase in online health misinformation, which, to a significant extent, has impacted the public's confidence and trust in online healthcare information. The World Health Organization (WHO) has termed this issue as the COVID ‘infodemic. In an effort to address this COVID 'infodemic', various datasets have been developed to facilitate the detection of misinformation. However, most of these datasets are limited to primarily unimodal data, which consists mainly of textual cues, and lack visual cues, such as images, infographics, and other graphic data components. Existing literature indicates that there are a few multimodal datasets to aid with COVID misinformation identification, but none of these have an organized, processed and analyzed repository of image artifacts in the form of visual cues. To address this gap, we introduce the novel CoVCues dataset, which contains various image artifacts and emphasizes the importance of visual cues for the detection of online health misinformation. CoVCues is a uniquely new framework that leverages categorization, sub-categorization of images from publicly available data, and AI models to explore the capability of visual content in enhancing misinformation detection. In order to demonstrate that CoVCues aids in establishing information assurance by fostering trust, we apply computer vision based machine learning (ML) techniques to find pattern recognition elements, such as identified faces, and image coherence plus authenticity, which can be connected directly to well-known trust antecedents. Through this applied trust analysis process, we make a strong case of information reliability for the trustworthiness of the CoVCues image dataset. Overall, our CoVCues dataset introduces a novel image-centric approach to empower misinformation detection accuracy and represents a valuable resource for both researchers and professionals in fighting against the COVID infodemic. CoVCues is a uniquely exclusive image artifacts collection to help extend the value of the multimodal datasets in this context and could open new avenues of future research and development on this topic area. It highlights how visual cues can increasingly play a pivotal role in online health misinformation recognition, facilitating the development of more effective and robust detection methods for COVID related infodemic crises. Additionally, to our knowledge, our approach of organization, classification and mapping of the collected CoVCues image artifacts to critical trust factors using computer vision driven ML methods, as described in this paper, is novel and has no precedence.

This document is currently not available here.

Share

COinS
 

CoVCues: A Trustworthy Resource Amidst The COVID Infodemic

The COVID-19 pandemic led to an increase in online health misinformation, which, to a significant extent, has impacted the public's confidence and trust in online healthcare information. The World Health Organization (WHO) has termed this issue as the COVID ‘infodemic. In an effort to address this COVID 'infodemic', various datasets have been developed to facilitate the detection of misinformation. However, most of these datasets are limited to primarily unimodal data, which consists mainly of textual cues, and lack visual cues, such as images, infographics, and other graphic data components. Existing literature indicates that there are a few multimodal datasets to aid with COVID misinformation identification, but none of these have an organized, processed and analyzed repository of image artifacts in the form of visual cues. To address this gap, we introduce the novel CoVCues dataset, which contains various image artifacts and emphasizes the importance of visual cues for the detection of online health misinformation. CoVCues is a uniquely new framework that leverages categorization, sub-categorization of images from publicly available data, and AI models to explore the capability of visual content in enhancing misinformation detection. In order to demonstrate that CoVCues aids in establishing information assurance by fostering trust, we apply computer vision based machine learning (ML) techniques to find pattern recognition elements, such as identified faces, and image coherence plus authenticity, which can be connected directly to well-known trust antecedents. Through this applied trust analysis process, we make a strong case of information reliability for the trustworthiness of the CoVCues image dataset. Overall, our CoVCues dataset introduces a novel image-centric approach to empower misinformation detection accuracy and represents a valuable resource for both researchers and professionals in fighting against the COVID infodemic. CoVCues is a uniquely exclusive image artifacts collection to help extend the value of the multimodal datasets in this context and could open new avenues of future research and development on this topic area. It highlights how visual cues can increasingly play a pivotal role in online health misinformation recognition, facilitating the development of more effective and robust detection methods for COVID related infodemic crises. Additionally, to our knowledge, our approach of organization, classification and mapping of the collected CoVCues image artifacts to critical trust factors using computer vision driven ML methods, as described in this paper, is novel and has no precedence.