Graduate Student Talks -- Ottawa Room (2nd floor)
11:00 - 11:50 AM Saturday October 22nd
​

Approximate Time	Session Title	Speaker
11:00	Mapping Urban Mobility: A design investigation into spatial justice in Toronto In this paper, we combine critical cartography, civic technology, and human-computer interaction (HCI) to enable participation from climate activists in the planning and execution of Toronto’s active travel infrastructure. This effort describes qualitative research methods to create a digital map of climate action resources and climate citizens, economies and infrastructure in Toronto that work to address vulnerabilities introduced by climate change in the community. We seek to include organizations that advocate for equitable policy goals and sustainable cities. This work is motivated by the following research questions: 1. How does the introduction of technology challenge existing models of climate justice in urban cities? 2. How can we design HCI interventions to enable community-driven climate activism in order to achieve equitable and sustainable cities for active travelers (cyclists, pedestrians, etc)? Our digital artifact uses counter-mapping, which refers to critical cartographic and feminist data visualization practices to render visible perspectives that are excluded from the current knowledge regime. We hope to create a map that (1) narratively maps amenity-rich and walkable communities, (2) expedites mobility planning and execution, and (3) serves as a living database to record and understand people’s relationship to space, in Toronto.  This design-based investigative project seeks to (1) advance a deep, situated understanding of computing technology’s role when engaging across multiple sites of climate advocacy work, (2) demonstrate a personalized, practical and spatial landscape of  transit planning, and (3) connect resources, identities, and issues for everyone working towards TransformTO. Today, computing faces challenges in building equitable, bottom-up participatory human intelligence. We hope to address these challenges by repositioning earlier civic technologies towards facilitating digitally supported advocacy and translating technology-mediated climate justice into real-world sustainable cities. Ultimately, these research questions and our work answer a broader series of questions in HCI around the relationship between social justice and social computing.	Taneea S Agrawaal University of Toronto ​
11:10	How data brokers endanger privacy In the recent years, public concern about the information trading industry is growing in importance as more and more data become available online. Indeed, information traders such as data brokers have increasing access to private information, enabling them to construct profiles of individuals with more accuracy and detail. As those profiles might contain sensitive information, data leakage would endanger customers’ privacy, putting them at risk of frauds or identity thefts. Even without data leakage, important amount of personal information is available for free on person search sites, which list a public registry of personal profiles they possess. Furthermore, as more data brokers enter the market, more private personal data become publicly accessible. Thus, it becomes easier to link data from different sources to build more complete profiles of an individual containing more sensitive information. To provide an understanding of the current data broker industry, we conducted a survey on 75 data brokers. Also, to highlight the danger of person search sites, we implemented a system which automatically collects and links profiles from different person search sites, constructing highly informative profiles about a specific individual. This system, named DROPLET, produces rich linked profiles, and is easy to use as it requires limited human intervention. ​	Muxue Guo University of Montreal
11:20	EXMULF: An Explainable Multimodal Content-based Fake News Detection In this work, we present an explainable multimodal content based fake news detection system. It is concerned with the veracity analysis of information based on its textual content and the associated image, together with an Explainable AI (XAI) assistant. To the best of our knowledge, this is the first study that aims to provide a fully explainable multimodal content-based fake news detection system using Latent Dirichlet Allocation (LDA) topic modeling, Vision-and-Language BERT (VilBERT) and Local Interpretable Model-agnostic Explanations (LIME) models. Our experiments on two real-world datasets demonstrate the relevance of learning the connection between two modalities, with an accuracy that exceeds 10 state-of-the-art fake news detection models.	Dorsaf Sallami University of Montreal ​
11:30	Explainable AI for Clinical Decision Support: Literature Review, Key Gaps, and Research Synthesis Background: Clinical decision support (CDS) systems are intended to facilitate the process of decision making in clinical settings. Artificial intelligence (AI) methods, particularly in the field of machine learning (ML), have the potential to augment CDS systems by synthesizing complex relationships in data into predictions that are relevant to the decision at hand. However, the black box nature of many ML models hinders their applicability in this context, because decision-makers need to understand the rationale behind these machine predictions to be able to trust them. The general challenge of explaining the workings of AI models has sparked research interest in explainable AI (XAI). XAI methods are used to construct and communicate explanations of how a model operates and may offer a way of increasing the applicability of AI methods to CDS systems. Objective: This study aims to identify, summarize, and evaluate the available research, current state of utility, and challenges of applying XAI methods in CDS systems. Methods: PubMed, CINAHL, PubMed, IEEE, ACM, Web of Science, and Google Scholar were searched with keywords related to “XAI”, “clinical decision support,” and “machine learning”. The results were then restricted to those that addressed CDS systems and XAI. We analyzed the resulting literature to identify the current state of the art of applying XAI methodology to CDS systems, and we identified key gaps. Results: There were 34 studies that met our criteria. We identified two gaps in the application of XAI in clinical decision support: first, existing XAI methods are rarely informed by end-user needs and second, user evaluations of model interpretability are lacking. By approaching decision support from the foundation of human cognitive activities, we propose a conceptual framework for the design of a user centered approach for the development of XAI for CDS systems, and we discuss implications for XAI evaluation in the context of CDS systems. Conclusions: We identified important gaps in the design of XAI methods for CDS systems. We proposed a framework to bridge the gap between XAI methods and end-users by analyzing how XAI methods should be incorporated into decision-making by framing it as cognitive activity. We also identified that interdisciplinary research teams which include clinicians, cognitive scientists, and computer scientists are needed to advance XAI for CDS systems by producing user-centered explanations.	Mozhgan Salimiparsa University of Western Ontario ​
11:40	Challenges in coding for visually impaired in India In India, and probably in many developing countries, there are not enough accessible resources available to teach computer science and programming to the visually impaired (VI). During our primary research at Perkins School For the Blind, Boston, three different solutions, CodeJumper, CodeQuest and Quorum were identified which are used by the visually impaired for coding. We compared these solutions through the lens of factors like accessibility, affordability, language feasibility, engagement, usability and final outcome. Simultaneously, our findings from indian blind schools concluded that coding is almost non existent for VI in indian blind schools. All three tools were analysed and compared and their relevance to indian blind students was extrapolated.  Right from factors like costing and language, to the training of teachers, there are various improvements needed and there is a long way to go. To fill the gaps, solutions like audio translators, crowdfunding, training of teachers require collaborative efforts of stakeholders like institutions, creators of technology, parents of VI students, and teachers as well. Only then can these existing technologies be spread out to other countries and can be considered universal solutions.  More active initiatives and interventions are needed at ground level to make a robust computer science curriculum at blind schools in India so that coding and programming can be an integral part of their pedagogy. Existing technologies may need to be tweaked and modified to make it relevant to VI students in India and the third world.	Ananya Jain University of Toronto ​