Concerns about the negative impact of generative artificial intelligence (GenAI) on post-secondary learning continues to drive vital discussions. University of Calgary professor Soroush Sabbaghen is encouraging his peers to look past the stigma and embrace GenAI as a highly customizable tool that can be harnessed to improve student success.
Sabbaghen developed a tool called SMARTIE — a Strategic Module Assistant for Rubrics, Tasks and Inclusive Education. This first-of-its-kind suite of web-based applications works like an intelligent assistant, enabling educators to design inclusive course components and by extension, create learning environments that are accessible and adaptable for all students.
“There are a lot of faculty who dislike GenAI, but it’s actually a tool that enables us to set students up for success in unprecedented ways,” said Sabbaghan, who is also director of Equity, Diversity and Inclusion (EDI) at the Werklund School of Education.
Sabbaghan, who embraced the power of the GPT-4 application programming interface to build SMARTIE, thinks of GPT-4 as the enhanced next level of ChatGPT.
Sabbaghan will share the impact of SMARTIE at the upcoming Congress of the Humanities and Social Sciences (Congress 2024), Canada’s largest academic gathering and one of the most comprehensive in the world, taking place June 12 to 21 at McGill University in Montreal.
Integrating EDI principles into current teaching practices creates an enormous cognitive burden and is extremely time-consuming for teachers. SMARTIE streamlines the process by providing recommendations along with tailored support in a matter of seconds. Since making the tool freely available online, Sabbaghan has recorded more than 5,000 hits.
Sabbaghan in consultation with faculty, then tested and refined SMARTIE to remove variability and improve accuracy. The program currently supports multilingual students as well as students with dyslexia and visual or auditory impairment. There are also plans to support neurodiversity.
Sabbaghan will demonstrate how different faculties are using the technology to achieve game-changing results like professors in the University of Calgary’s Faculty of Engineering using SMARTIE to measure the diversity of their reading lists to ensure marginalized groups are represented and that faculty are meeting their goals for theoretical, multidisciplinary and temporal diversity.
Meanwhile, members of the Faculty of Social Work are finding SMARTIE useful for personalizing tasks, developing fair assessment strategies and ensuring experiential learning activities are equitable.
To date, Sabbaghan has released seven apps aimed at assisting with course structure, learning activities and rubric design. Unlike other GenAI tools that search the web broadly, SMARTIE is linked to trusted inclusivity sources only and apps are rigorously vetted before being released.
“This is about moving in the direction of creating courses that allow everyone to succeed, regardless of their background, personality or ability,” said Sabbaghan.
Plans are underway to develop a version of the tool suitable for kindergarten to grade 12 learning environments as well.
“The idea is to help you think, not to copy and paste. We don’t want teachers to rely on AI. We want it to give them recommendations so they can make their own judgements and decisions,” he said.
Another AI marketexpanding exponentially is healthcare with the global market forecast to reach $88 billion US by 2028. Yet, without input from the people using these technologies, designers may engineer systemic biases and harmful stereotypical assumptions into these apps.
A first-of-its-kind study, Towards an Accessible and Inclusive A.I. (AI2), is calling for a paradigm shift in the way AI developers approach solutions for people with disabilities.
The leads on the study include Christo El Morr, York University Professor of Information Technology; Rachel Gorman, York University Associate Professor of Critical Disability Studies; and University of Calgary Associate Professor, Faculty of Social Work, Yahya El-Lahib,
AI2 research assistant, Bushra Kundi, will be sharing the teams’ findings as a featured speaker at Congress 2024. Kundi is a research assistant at York University; Master of Science student in e-Health at McMaster University; and a Senior Clinical Applications Specialist at Oakville-based Nucleus Independent Living.
“The problem with current health AI tools aimed at assisting people with disabilities is that they only focus on the medical model, which sees disability as a problem and looks at helping someone overcome their limitations,” said Kundi.
AI2s research found a need for a more inclusive approach to achieve accessible and inclusive AI. To achieve that goal means looking at the social model of disability, including environmental changes that can be implemented to meet needs, and that requires a more ‘disability-centric’ AI design process with input from multi-disciplinary teams as well as people living with disabilities.
“If we don’t shift to a more inclusive and collaborative methodology, we run the risk of producing solutions that actually perpetuate societal biases against people with disabilities,” Kundi explained.
At Congress 2024, Kundi will highlight how AI is currently being applied to help people with disabilities. Kundi’s work revealed that despite AI’s potential to support self-management of health conditions, enhance assistive devices and further disability justice, models developed to date have not sufficiently measured or addressed bias.
“Essentially, an ableist perspective prevails within AI, potentially exacerbating disparities rather than alleviating them,” said Kundi.
That’s because historical data used to train AI systems can be inherently prejudiced against people with disabilities and is often based on beliefs that typical abilities are superior. Even generative AI tools like ChatGPT, which are rapidly opening up opportunities to help people with speech or hearing disabilities, could have biases.
“We need to ensure that we’re considering people with disabilities and the social model of disability so that we design inclusive AI solutions. Software engineers are skilled in creating AI models, but without that support and insight from the person living with the disability, they will never be able to cover all of the bases required to make their model disability-centric,” said Kundi.
Based on their findings, the researchers are calling for a collaborative effort to reorient AI development towards a disability-focused framework and are urging developers to include the user perspective when creating data sets and AI models.
“If we want to ensure equitable benefits of AI advancements for all members of society, it is imperative that AI systems advance beyond technical excellence to encompass social responsibility and inclusivity as well.”
At Sustaining shared futures — The Federation for the Humanities and Social Sciences Congress 2024 — thousands of research papers will be unveiled and presentations made by social sciences and humanities experts from around the globe.
More than 8,000 scholars, graduate students and practitioners are expected to participate in the ten-day event focusing on generating solutions for today while creating the sustainable systems of tomorrow and ensuring the inclusion of inspiring ideas, dialogue and action that create a more diverse, sustainable, democratic and just society.
Organized by the Federation for the Humanities and Social Sciences in partnership with McGill University, Congress 2024 is sponsored by the Social Sciences and Humanities Research Council of Canada, Universities Canada, the Canada Foundation for Innovation, University Affairs, Sage, and The Conversation Canada.
Visit Congress 2024 to register for a community pass and access the program of events open to the public from June 12 to 21, 2024. Registration – which includes 140+ keynote and open Congress.