Artificial Intelligence for Students in Postsecondary Education

AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020

Artiﬁcial Intelligence for Students in Postsecondary Education: A World of Opportunity Natalina Martiniello (University of Montreal´ (School of Optometry), Centre for Interdisci- plinary Research in Rehabilitation of Greater Montreal´ (CRIR), Adaptech Research Network; [email protected]) Jennison Asuncion (Adaptech Research Network; [email protected]) Catherine Fichten (Adaptech Research Network, Dawson College, McGill University (De- partment of Psychiatry), Jewish General Hospital (Department of Psychiatry, Behavioral Psy- chotherapy and Research Unit); catherine.ﬁ[email protected]) Mary Jorgensen (Adaptech Research Network; [email protected]) Alice Havel (Adaptech Research Network, Dawson College; [email protected]) Maegan Harvison (Adaptech Research Network; [email protected]) Anick Legault (Adaptech Research Network, Dawson College; acle- [email protected]) Alex Lussier (Adaptech Research Network; [email protected]) Christine Vo (Adaptech Research Network; [email protected]) DOI: 10.1145/3446243.3446250

Abstract unique needs of users. While the practical and mundane benefits of artificial intelligence sys- AI-based apps can facilitate learning for all tems are taken for granted by society at large post-secondary students and may also be (e.g. speech recognition, real-time caption- useful for students with disabilities. Here we ing (Mathur, Gill, Yadav, Mishra, & Bansode, share some reflections from discussions that 2017), linguistic translation (Kapoor, 2020), or took place during two advisory board meet- organizing tools such as calendars and “to ings on the use of such apps for students with do” lists (Canbek & Mutlu, 2016)), often over- disabilities at the post-secondary level. looked are the potential benefits to specific Keywords: college and university students populations who would otherwise be depen- with disabilities, artificial intelligence apps, dent upon third parties for support. For exam- mobile AI apps ple, artificial intelligence has been success- fully harnessed to provide people with mental health concerns an always-available, point- Introduction in-time responsive, supplementary or interme- Intelligent technologies (such as smartphones diate support system (Inkster, Sarda, & Sub- and tablets which incorporate principles of uni- ramanian, 2018). The available evidence at versal design) have the potential to increase present indicates that individuals with disabili- the inclusion of students with disabilities and ties are conspicuously excluded from AI train- other diverse learners in different aspects of ing models and have not yet become mean- post-secondary education. Artificial intelli- ingful contributors to, or beneficiaries of, the gence (“AI”), which for our purposes includes ongoing discussions surrounding artificial in- “computing systems that are able to engage telligence and machine learning (Lillywhite & in human-like processes such as learning, Wolbring, 2020). adapting, synthesizing, self-correction and use of data for complex processing tasks” Despite advances in inclusive education prac- (Popenici & Kerr, 2017), is in many respects tices, the reality is that for many of the 10%- the bedrock of more universally accessible en- 20% (Eagan et al., 2017; Fichten et al., 2018; gagement, because it has the potential to per- Snyder, De Brey, & Dillow, 2016) of students mit technology to adapt to the diversity and with disabilities, barriers to accessing post- secondary education continue to persist. In- Copyright c 2020 by the author(s). accessible websites, commonplace at some of

17 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020 the world’s most renowned institutions, con- The Advisory Board Meetings tinue to pose significant accessibility barriers (Huffington, Copeland, Devany, Parker-Gills, We invited stakeholders from within our net- & Patel, 2020). In spite of increased use of work who could provide insights on the use automatic speech recognition, Deaf and hard of artificial intelligence at the post-secondary of hearing students often do not have ade- level from diverse perspectives. In total, this quate access to qualified real-time captioning included 38 individuals: 7 students, 3 disabil- (Butler, Trager, & Behm, 2019). Even where ity/accessibility service providers, 14 faculty captioning or interpreters are available, Deaf members (some with and some without dis- and hard of hearing students may lag be- abilities), 9 technology experts, and 5 tech- hind their peers due to the increased cognitive nology users with disabilities. Two advisory load of processing the visual translation of au- board meetings were held in May 2020, us- dio information, visual attention limits associ- ing the Zoom (San Jose, CA) videoconferenc- ated with trying to track dispersed visual infor- ing system to accommodate for the different mation in the classroom, difficulties engaging time zones in which the international atten- in discussions, and limited social interactions dees resided. Spontaneously, attendees di- with peers, among other factors (Kushalnagar, vided themselves roughly in half into one of 2019). the two sessions. The following overarching questions were distributed in advance of the The initiative we report in this article had its meetings and provided a framework for the genesis in prior research exploring the is- discussion: sue of fairness of AI for people with disabilities (in employment, education, public safety, and healthcare (Trewin et al., 2019)). 1. What AI-based smartphone or tablet tech- Building on our previous research regard- nologies have you come across that are cur- ing the accessibility of information and com- rently used by college and university stu- puting technologies for post-secondary students with disabilities? Which ones work dents (Fichten, Asuncion, & Scapin, 2014; well? Which ones do not? Thomson, Fichten, Havel, Budd, & Asun- cion, 2015) and the increasing use and util- 2. What AI-based smartphone and tablet tech- ity of ubiquitous mobile technologies, such as nologies are out there but are rarely consid- smartphones and tablets in post-secondary ered even though they could help college education (Fichten et al., 2019), the ques- and university students with disabilities do tion arose as to the impact and adoption academic work? of new AI-based technologies. We there- fore convened an advisory group to gather input from post-secondary students, con- 3. What are some functions you can foresee sumers with disabilities, post-secondary dis- in the future that AI-based smartphone and ability/accessibility service providers, faculty, tablet technologies could perform to assist and technology experts to explore how college students with disabilities succeed in higher and university students with diverse disabili- education? (i.e. It would be great if an ties (visual or hearing impairments, learning AI-based smartphone and tablet technology disabilities, attention deficit hyperactivity disor- could help a student do X.) der, etc.) could benefit from AI-based smartphone and tablet apps, and how these could facilitate student academic success. In sum- Summaries of comments from members of the marizing the themes that emerged during the advisory board, including the identification of advisory board meetings, we hope to lay the specific tools and resources (see Adaptech groundwork for more in-depth and targeted ini- Research Network, 2020), were then pre- tiatives to objectively evaluate the effective- pared. The resulting summaries, grouped and ness of the tools currently being used, and to categorized by Adaptech Research Network explore the feasibility and potential success of team members according to theme and the future innovations within the post-secondary nature of the application, are described more education context. fully below.

18 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020

Advisory Board Meeting Themes significant disabilities (e.g. those who are functionally blind) may actually develop more General observations sophisticated technological capabilities than Those who need AI the most are also the their peers. For example, Fichten, Asuncion, most likely not to be able to use AI: A Barile, Ferraro, and Wolforth(2009) found that recurring comment was that systems relying students who were functionally blind felt sig- on artificial intelligence have the potential to nificantly more comfortable utilizing technol- open the world to individuals who cannot ac- ogy in the classroom than students who had cess information or environments in the same low vision. Nonetheless, many instances were manner as everyone else. Unfortunately, be- noted whereby students who could benefit cause AI is based on training data and learn- from the use of AI-enabled technologies (such ing analytics derived from a generally able- as Seeing AI) were simply unaware that such bodied population, those who veer from what tools even existed. Learning that the technolo- is viewed as ’the norm’ and who have distinc- gies exist and how to effectively use them is, tive voice, facial, or movement characteristics for many, a critical step in the transition from are likely to be misunderstood (or not under- high school to post-secondary education. The stood at all) by AI capture processes. As one lack of training is especially problematic for individual described it: students who acquire a disability later in life and who do not have previous experience with AI programs are made up of algorithms, the use of such devices. or a set of rules that help them identify Similarly, many students are unaware of patterns so they can make decisions with the convenience and power that could be little intervention from humans. But algo- achieved by leveraging the integrated use of rithms need to be fed data in order to learn AI-enabled technologies. For example, the those rules - and, sometimes, human prej- online service “IFTTT” (If-This-Then-That) al- udices can seep into the platforms. lows users to schedule and pre-program cas- cading events based on triggers from web- Academic discussions surrounding the ethics enabled services. IFTTT could be used to de- of AI implementation are focused on issues tect (based on GPS data from a smartphone surrounding the potential for discrimination or or a Fitbit) that a user is not in the class- exclusion based on gender, race, language, room where they are expected to be accord- or other identifiable characteristics. The is- ing to their calendar and trigger a notification sues of AI bias in the disability context go be- to make them aware that they are missing a yond these discussions because, unlike race class. or gender, there is no uniform cluster or data element that is common to everyone with a disability. Whether a particular AI integration Current and potential utilization of will be useful may, for some users, depend existing AI tools to facilitate on the degree to which additional idiosyncratic characteristics can be taken into account. In post-secondary learning most applications, that functionality does not Chatbots exist, and current research shows that the training data used to “teach” AI systems is “Chatbots” – AI-enabled text-based conver- lacking in diversity by not including individuals sationalists – are becoming common as a with disabilities (Kafle et al., 2020). means of providing front-line support to users of technology and have the ability to provide Students with disabilities may be more quick answers to the most common ques- technologically advanced than their peers, tions that users might otherwise pose to a but greater training and support is re- technical support agent. Chatbots have also quired: Students with disabilities who face been implemented in post-secondary institu- daily barriers to access often develop a height- tions to provide on-demand answers to the ened ability to problem-solve and this can many questions that students pose. Vari- manifest through the use of technology. Past ous universities have used AI-enabled Chat- research has shown that individuals with more bots to facilitate the transition from high school

19 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020 into university, guiding new students through assistant that helps with making decisions, the myriad of administrative matters that must managing symptoms of anxiety, and re- be addressed at different points of the ad- sponding to unexpected situations. Once missions cycle, such as applying for and re- the individual subscribes to this application, sponding to financial aid applications and en- they gain access to a personal specialist to rolling/registering for courses at the appro- assist in setup, accessible self-management priate time (Nurshatayeva, Page, White, & tools, and contact with a human support net- Gehlbach, 2020; Page & Gehlbach, 2017). work. In each case, the implementation of an AI- • Empower Me is a digital coach that operates enabled chatbot was found to improve out- on smart glasses, to aid individuals with comes for first-year students, especially for autism in self-regulating by helping them those who were struggling or particularly ’at to understand facial expressions and emo- risk’. Aside from being responsive to students’ tions. Where appropriate, it draws attention point-in-time needs, these Chatbot services to facial and eye cues. also permitted admissions and financial aid staff to redeploy resources to the more com- • SeizAlarm, My Medic Watch, and Smart- plex cases. Building on these past successes, Watch Inspyre can detect seizures and several meeting attendees noted that Chat- trigger notifications to emergency contacts, bots are also being implemented at their own providing information on the user’s location institutions, primarily to support the needs of and current status. distance education students. These are being • Woebot is a mental health “chatbot” which integrated into existing learning management can provide an outlet for students with systems (e.g. Moodle) to provide answers to depression, and has been shown to re- common student questions such as, “When is duce depressive symptoms by employing my exam?” or “How can I organize my course positive thinking precepts commonly used documents?” in traditional Cognitive Behavioral Therapy (Fitzpatrick, Darcy, & Vierhile, 2017). Emotional, mental health, and medical regulation Many wearable technologies, such as “smart watches”, can also cue in to increased lev- Students who experience emotional hurdles els of stress or anxiety, and provide reminders or mental health difficulties, many of which to take breaks and focus on breathing exer- can be episodic in nature, are at a distinct cises in response. These tools cannot sup- disadvantage in the post-secondary environ- plant the intervention of professional coun- ment which has traditionally lacked the flexi- selling, but may be useful in times of crises bility to adapt to varying or changing individ- or when more structured support services are ual needs (Arim & Frenette, 2019). Meaning- not readily available (Inkster et al., 2018). ful access to mental health support services has been found to be limited for many Cana- dian post-secondary students, despite the fact Organizational and executive functioning that mental health problems including anxiety, aids depression, and substance abuse are partic- One of the most common areas where AI- ularly acute among young adults age 18-25 enabled apps and tools were thought to be (Nunes et al., 2014), and are especially preva- useful as an educational aid is in the area lent during the current COVID-19 pandemic of personal organization and assisted execu- (Son, Hegde, Smith, Wang, & Sasangohar, tive functioning. More specifically, AI-enabled 2020). tools are being used to help coordinate the Where such support does not exist or is more many dependent and parallel tasks that stu- limited, several AI-informed app-based tools dents must complete. For example, while were identified which can provide point-in-time personal assistants such as Google Assistant support to aid in emotional, mental health, and are, of course, capable of giving reminders, medical regulation: such reminders are far more helpful if they are presented in a context-aware manner (Singh • Brain in Hand is an AI-enabled professional & Varshney, 2019). For example:

20 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020

• Aside from relying on items in an individual’s Accessing visual or textual information in calendar, AI can detect and recognize pat- alternative formats terns in an individual’s schedule and provide appropriate notifications, such as when an For students who are blind, have low vision, or individual should leave for a meeting. have other print disabilities such as a learning disability that impact the ability to read printed • For those commuting to and from work or information, tools that facilitate the conversion school, Google Maps will often learn one’s of text into more accessible formats (e.g. au- usual departure time and provide an alert in- dio) are particularly valuable. Text-to-speech dicating the expected travel time on a given systems that are capable of scanning a printed day as a subtle reminder of when to leave. page of text and reading it aloud to users date • Personal assistants, such as Siri and back to the 1970s, with the introduction of the Google Assistant, can be set to provide re- Kurzweil Reading Machines (Goodrich, Ben- minders at specific times or when reach- nett, De l’Aune, Lauer, & Mowinski, 1979). ing specific places (e.g. home, the grocery However, over the past decade, the prolif- store, etc.), when the user anticipates being eration of AI technologies has permitted the able to respond to those reminders. development of far more sophisticated “computer vision” applications, allowing apps to Meeting attendees noted that while some stu- identify or locate objects in the environment, dents are using Siri, Google Assistant, Alexa, describe scenes and photographs, and pro- and other AI-enabled apps, they are often vide real-time navigation assistance. unaware of the power and context-enabled options that could be used to provide more With respect to accessing textual information, meaningful and useful organizational assis- the team identified a number of existing “apps” tance. and technologies which facilitate this task, including:

Input mediation • Seeing AI or Office Lens, which use the cameras on smartphones and tablets to Entering information into a smartphone or have the text that is in front of the camera tablet can be difficult for students who have spoken aloud. physical or cognitive disabilities that impede interaction with the conventional keyboard. • Technologies that aid in summarizing long Several tools that mediate this interaction and pieces of text into a more abstract form. help to improve the speed and efficiency of Available technologies do exist, such as entering textual information were described by • SMMRY or Reddit’s AutoTLDR “bot”, and the team, including: research continues into making such automatic summaries more useful and accurate • SwiftKey and FlickType are AI-enabled key- (Dangovski, Jing, Nakov, Tatalovic, & Sol- board applications that learn and adapt to jacic, 2019). match an individual’s unique way of typing. • OrCam, a voice-activated wearable technol- • UNI is an AI-enabled device (which requires ogy that uses a small onboard camera to a subscription service) that converts tex- read text. tual information into sign language and vice • Voice Dream Scanner, an optical character versa, including the ability to define custom recognition tool that can use the camera of and unique signs where required. a smartphone or tablet to read short texts as • Word prediction (which uses AI to under- well as longer multi-page documents. stand context and recall common words • SensusAccess provides a self-service solu- used by a writer) is a built-in feature in most tion to convert a range of electronic formats smartphone and tablet devices. into electronic braille, audio MP3, DAISY, • Language processing AI applications, and and e-book formats. linguistic revision tools such as Antidote, provide context-aware writing and revision Importantly, while the above tools are all assistance, which is particularly helpful for specialized programs targeted at those with students with learning disabilities. specific disabilities, use of the text-to-speech

21 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020 functionalities of modern smartphones, smart acute. There are, however, others who may watches, and smart speakers can be used by also benefit from access to information in non- any student who would benefit from bi-modal auditory forms, including those with auditory learning, or from effective proofreading, where processing deficits, as well as students who they can listen to the text while also reading it. simply receive and retain information more ef- It is important to note, however, that the ac- fectively when it is in a written form. With curacy of such applications will depend on the the widespread and mainstream adoption of quality of the text that is being scanned. smart speakers and smartphone-based personal assistants (e.g. Amazon Alexa, Google For accessing non-textual information, the fol- Home, Apple Siri, Samsung Bixby), a signif- lowing tools have been utilized: icant amount of interest has been generated • Seeing AI, which uses the cameras on toward improving the quality and accuracy of smartphones and tablets to identify objects speech recognition technologies. However, as or scenes in front of the camera. described by in Rabiner and Juang(2008), this remains true today: • OrCam, a voice-activated wearable technology that uses a small onboard camera The quest for a machine that can rec- to recognize and identify known faces, and ognize and understand speech, from any which can also identify products based on speaker, and in any environment has their universal bar code. been the holy grail of speech recognition • CamFind, which can identify objects in a research for more than 70 years. Although photo or video stream. we have made great progress in under- • Facebook and Twitter now use artificial in- standing how speech is produced and telligence to automatically provide image analyzed, and although we have made captions for photos and pictures. The qual- enough advances to build and deploy in ity of these descriptions is improving, and the field a number of viable speech recog- this is important for all students given the nition systems, we still remain far from the increasing frequency with which instructors ultimate goal of a machine that communi- and academic institutions are using social cates naturally with any human being. media to communicate. However, it should be noted that manual alt text descriptions The team described a wide range of means by should continue to be incorporated to en- which alternative representations of speech sure that such images are accessible to can be automatically generated: screen-reader users. • Public services such as YouTube and For students who may be learning in their Google Slides now provide or enable the second language (or learning a second lan- use of automatically generated captions, al- guage), translation tools such as Google though the accuracy is very much depen- Translate, Microsoft Translator, or DeepL can dent on the sound quality of the original pro- also facilitate access to information. AI has in- duction and varies significantly across lan- creased the speed and accuracy of automated guages. translation tools, but there remain significant • Some institutions maintain subscriptions limitations on these tools and the impact of with external service providers to generate context and nuance in language make it diffi- automatic captions on pre-recorded videos. cult to rely and trust in automated translations (Pantea, 2019). • The Zoom videoconferencing service allows for real-time captioning by a meeting host, or through connections to external captioning Accessing auditory information in providers (as well as enabling use of auto- non-auditory formats matically generated captions). WebEx and For students who are Deaf or who are hard Microsoft Teams have similar features. of hearing, the need to access information • Most smartphone and tablet devices include from classroom (or online) lectures and videos a built-in feature to “dictate” (rather than type through sign language or a textual format is text) through their virtual personal assistant,

22 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020

both to send a short message and for writ- the ability to clearly vocalize and articulate ing longer content (e.g. an email message, words and sounds. However, some individ- Office 365 Word document). uals suggested that, given the learning ca- • Just Press Record is an AI-enabled mobile pability of these systems, those experienc- audio recording “app” that permits recording more gradual changes in their vocaliza- ing, transcription, and iCloud synchroniza- tion ability may afford the AI an opportunity tion of voice memos across iOS devices. to learn these changes over time, and there- fore not experience the same degree of dis- • Translation tools such as Microsoft Transla- connect. Moreover, specialized speech recog- tor or Google Translate can also be used nition applications such as Voiceitt (formerly to generate a “real time” transcript if the known as Talkitt), which are designed specif- speaker is wearing a microphone. This per- ically to learn and understand non-standard mits students to both listen to the presenta- speech patterns, are being actively developed tion and follow along with the text. (Ibrahim, 2016). Several individuals also noted that the above tools could be used by any student who might, Navigation and environmental exploration for example, want to talk through a plan, cre- tools ate an outline, or simply brainstorm ideas, Finally, the role of AI-enabled GPS and nav- without becoming fixated on the task of writ- igation “apps” was discussed, largely in re- ing. spect of individuals who are blind or have Concerns over privacy and security have led low vision, for whom the ability to travel inde- some to eschew the use of these built-in pendently has been associated with greater speech-to-text features (which rely on cloud- self-confidence, stronger employment out- based technologies, where audio is sent to a comes, and more successful independent liv- remote processing center for analysis, and the ing (Vaughan, 1993). Several AI-enabled tools text returned to the user’s device) in favor of were described as being of assistance to inde- “on-device” solutions (such as Dragon Natu- pendent travel, including: rally Speaking), even though such technologies are less advanced and have a lower over- • AIRA (Artificial Intelligence and Remote Ac- all accuracy. cess), a subscription-based service that pairs AI-enabled analysis with a sighted A separate concern was identified regarding assistant (using the video camera on a the linear nature of audio recordings, and the smartphone) to guide individuals through inherent difficulty for anyone in locating spe- the physical environment. cific information within such a recording. It was suggested that AI and speech recogni- • BlindSquare, an AI-enabled GPS navigation tion may provide an answer to this problem app that provides outdoor navigation assis- by permitting a user to “search” audio for spe- tance, with the feedback, amount and na- cific words. While no specific application or ture of information being tailored to individ- technology was discussed, this approach is ual user needs and that which is most likely actually being used and further developed to to be useful at a given point in time. predict positive COVID-19 test results based • Microsoft Soundscape uses AI to enable on audio recordings (and key words such as navigation through the physical environment smell and taste) from telemedicine assess- using a series of 3D audio cues. ments (Obeid et al., 2020). • Nearby Explorer Online is a free GPS ap- Consistent with existing empirical evidence plication that helps students who are blind (e.g. Rohlfing, Buckley, Piraquive, Stepp, navigate the physical environment. Specific & Tracy, 2020), it was noted by individu- or favorite locations can be marked to pro- als that mainstream speech recognition tech- vide more contextual information. nologies do not work especially well for those with a voice or speech disorder, those For other users, AXS Map uses artificial intel- who have strong accents, or who have any ligence and crowdsourced accessibility infor- other condition which significantly impacts on mation to help locate accessible businesses

23 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020

(with wheelchair ramps) and nearby accessi- apps need to be retrofitted for accessibility. ble washrooms. AI developers often set out to address a need within a targeted population (e.g., tools that Future AI-based smartphone and tablet facilitate the conversion of text into more ac- technologies cessible formats for students who are blind or have low vision) without recognizing that there Probing exhibited a variety of single re- could be a broader application. Awareness of sponses to novel ways to use AI; their fre- this may not only increase the number of end- quency is too low to report here. However, users who can benefit, but funding for product what came up more frequently were calls to development may be easier to obtain. improve upon existing applications of AI. By far, improving and optimizing AI for caption- When it comes to AI based apps, develop- ing/live transcriptions/notetaking was raised ers should underscore the need for accessi- most as holding more promise to assist post- ble training documents, including those avail- secondary students with disabilities. The sec- able through YouTube and Google, to sensi- tion “Accessing auditory information in non- tize post-secondary students to both the ex- auditory formats” above provides an excellent istence and the potential of AI based mobile view of the current state of this area. While apps that can help them in their learning. auto-captioning exists, so has the popularity Finally, students with disabilities should be of the term Craptions, to refer to the fact that viewed as valuable stakeholders in the area while machines can produce captions based of AI development. Not only can they suggest on speech recognition, their quality is still not the need for the development of AI apps for where it needs to be. The #NoMoreCraptions novel uses, but also what improvements are Campaign (3PlayMedia, 2019) speaks to this needed for existing applications of AI. problem. The other reoccuring comment relative to im- Conclusion proving existing AI use related to AI for as- sisting with organization/routine setting. This Assembling an advisory board of stakeholders does not come as a surprise given that the representing students, disability/accessibility largest numbers of students with disabilities service providers, faculty members, technol- have cognitive, attention deficit or learning dis- ogy experts, and technology users with disabilities where organization and structure are abilities from five countries resulted in an ex- among the factors largely critical to their suc- tensive repository of information regarding AI cess (Fichten, Havel, Jorgensen, Arcuri, & Vo, apps, not only in terms of what’s out there that 2020). is being used, but how it is being used and by whom. Along with this comes advice about what AI apps would be welcome in the future Implications for AI Developers and which existing apps need to be upgraded. There are numerous implications from this dis- University and college students with disabil- cussion. First, AI developers need to include ities make up a large proportion of post- individuals at the “edges” and not just those secondary students. Twenty years ago, we that fit the dominant section under the nor- noted a trend for technologies intended for mal curve (Treviranus, 2018). In the case of non-disabled students to be adopted – and post-secondary students with disabilities, this used in novel ways - by students with disabil- means that studies are carried out in an ac- ities (Fichten, Barile, Fossey, & De Simone, cessible manner using oversampling of indi- 2000). More recently, we noted a similar trend viduals with different disabilities. Also, stu- for the cross-use of technologies intended for dents with disabilities need to be included non-disabled students to be used by students in the design of AI apps from inception, as with disabilities (Fichten et al., 2014). Now we proposed by the universal design paradigm have taken our first steps to gather compara- (Story, Mueller, & Mace, 1998, Chapter 3). ble information on the use of artificial intelli- This results in less work - and less expensive gence apps by college and university students work - down the line when poorly designed with and without disabilities.

24 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020

Acknowledgments Eagan, K., Stolzenberg, E. B., Ramirez, J. J., Aragon, M. C., Suchard, M. R., This project was made possible by a grant & Hurtado, S. (2017). The Ameri- from Le Poleˆ montrealais´ d’enseignement can freshman: National norms fall 2014 superieur´ en intelligence artiﬁcielle (PIA). (Vol. 36). Los Angeles: Higher Education Research Institute, UCLA. https:// www.heri.ucla.edu/monographs/ References TheAmericanFreshman2016.pdf. Fichten, C. S., Asuncion, J., & Scapin, 3PlayMedia. (2019, February). #NoMore- R. (2014). Digital technology, learn- Craptions campaign calls for bet- ing, and postsecondary students ter CCs on YouTube videos. with disabilities: Where we’ve been https://www.3playmedia.com/ and where we’re going. Journal of blog/nomorecraptions-campaign Postsecondary Education and Dis- -calls-for-better-ccs-on ability, 27(4), 369-379. http://www -youtube-videos/. .ahead-archive.org/uploads/ Adaptech Research Network. (2020). publications/JPED/JPED%2027 4/ Canadian and international experts JPED27 4 FULL%20DOCUMENT.pdf. weigh in: An annotated list of AI- doi: 10.1177/0145482X0910300905 related resources for college and Fichten, C. S., Asuncion, J. V., Barile, M., Fer- university students with and without raro, V., & Wolforth, J. (2009). Accessi- disabilities. https://adaptech bility of e-learning and computer and in- .org/wp-content/uploads/ formation technologies for students with AIurlsFromAIMeeting.docx. visual impairments in postsecondary ed- Arim, R., & Frenette, M. (2019). Are mental ucation. Journal of Visual Impairment & health and neurodevelopmental condi- Blindness, 103(9), 543-557. tions barriers to postsecondary access? Fichten, C. S., Barile, M., Fossey, M., & (Tech. Rep.). Ottawa, Canada: Statistics De Simone, C. (2000). Access to Canada. https://www150.statcan educational and instructional computer .gc.ca/n1/pub/11f0019m/ technologies for postsecondary students 11f0019m2019005-eng.htm. (Ana- with disabilities: Lessons from three lytical Studies Branch Research Paper empirical studies. Journal of Edu- Series – Catalogue no. 11F0019M – No. cational Media, 25, 179-201. doi: 417) 10.1080/1358165000250303 Butler, J., Trager, B., & Behm, B. (2019). Fichten, C. S., Havel, A., Jorgensen, M., Ar- Exploration of automatic speech recog- curi, R., & Vo, C. (2020). Is there an nition for deaf and hard of hearing app for that? apps for post-secondary students in higher education classes. students with attention hyperactivity dis- In The 21st International ACM SIGAC- order (ADHD). Journal of Education CESS Conference on Computers and Training Studies, 8, 22-28. doi: and Accessibility (pp. 32–42). doi: 10.11114/jets.v8i10.4995 10.1145/3308561.3353772 Fichten, C. S., Jorgensen, M., Havel, A., Canbek, N. G., & Mutlu, M. E. (2016). On the King, L., Lussier, A., Asuncion, J., . . . track of artiﬁcial intelligence: Learning Amsel, R. (2018). Information and with intelligent personal assistants. Jour- communication technologies: Views of nal of Human Sciences, 13(1), 592-601. canadian college students and ”excel- doi: 10.14687/ijhs.v13i1.3549 lent” professors. Journal of Education Dangovski, R., Jing, L., Nakov, P., Tatalovic, and Training Studies, 6(9), 1–12. doi: M., & Soljacic, M. (2019). Rotational 10.11114/jets.v6i9.3390 unit of memory: a novel representation Fichten, C. S., Jorgensen, M., King, L., Havel, unit for rnns with scalable applications. A., Heiman, T., Olenik-Shemesh, D., & Transaction of the Association of Com- Kaspi-Tsahor, D. (2019). Mobile tech- putational Linguistics, 7, 121-138. doi: nologies that help post-secondary stu- 10.1162/tacl a 00258 dents succeed: A pilot study of cana-

25 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020

dian and israeli professionals and stu- ternational ACM SIGACCESS Confer- dents with disabilities. International Re- ence on Computers and Accessibility search in Higher Education, 4, 35-50. (p. 569–571). New York, NY, USA: As- doi: 10.5430/irhe.v4n3p35 sociation for Computing Machinery. doi: Fitzpatrick, K. K., Darcy, A., & Vierhile, M. 10.1145/3308561.3354640 (2017). Delivering cognitive behavior Lillywhite, A., & Wolbring, G. (2020). Cov- therapy to young adults with symptoms erage of artificial intelligence and ma- of depression and anxiety using a fully chine learning within academic litera- automated conversational agent (woe- ture, Canadian newspapers, and Twit- bot): a randomized controlled trial. JMIR ter tweets: The case of disabled peo- Mental Health, 4(2), Article e19. doi: ple. Societies, 10(1), article 23. doi: 10.2196/mental.7785 10.3390/soc10010023 Goodrich, G. L., Bennett, R., De l’Aune, Mathur, P., Gill, A., Yadav, A., Mishra, A., W. R., Lauer, H., & Mowinski, L. (1979). & Bansode, N. K. (2017). Cam- Kurzweil reading machine: A partial eval- era2Caption: a real-time image caption uation of its optical character recognition generator. In 2017 International Con- error rate. Journal of Visual Impairment ference on Computational Intelligence in & Blindness, 73, 389. Data Science (ICCIDS) (p. 1-6). IEEE. Huffington, D., Copeland, B., Devany, K., doi: 10.1109/ICCIDS.2017.8272660 Parker-Gills, A., & Patel, J. (2020). As- Nunes, M., Walker, J. R., Syed, T., De Jong, sessing accessibility: Universal design M., Stewart, D. W., Provencher, M. D., for university websites. Disability Compli- . . . Furer, P. (2014). A national sur- ance for Higher Education, 25(10), 1–5. vey of student extended health insur- doi: 10.1002/dhe.30835 ance programs in postsecondary insti- Ibrahim, S. A. (2016). Talkitt app: tutions in Canada: Limited support for Changing the quality of life for peo- students with mental health problems. ple living with speech disabilities. Canadian Psychology, 55(2), 101-109. PACE University Student and Fac- doi: 10.1037/a0036476 ulty Research Days. Retrieved from Nurshatayeva, A., Page, L. C., White, C. C., http://digitalcommons.pace & Gehlbach, H. (2020). Proactive stu- .edu/sfresearchday/1 dent support using artificially intelligent Inkster, B., Sarda, S., & Subramanian, V. conversational chatbots: The importance (2018). An empathy-driven, conversa- of targeting the technology [EdWorking- tional artificial intelligence agent (Wysa) Paper no. 20-208]. doi: 10.26300/mp4q- for digital mental well-being: Real-world 4x12 data evaluation mixed-methods study. Obeid, J. S., Davis, M., Turner, M., Meystre, JMIR mHealth and uHealth, 6(11), arti- S. M., Heider, P. M., O’Bryan, E. C., & cle e12106. doi: 10.2196/12106 Lenert, L. A. (2020). An artificial intel- Kafle, S., Glasser, A., Al-khazraji, S., ligence approach to COVID-19 infection Berke, L., Seita, M., & Huenerfauth, risk assessment in virtual visits: A case M. (2020, March). Artificial intelli- report. Journal of the American Medi- gence fairness in the context of accessi- cal Informatics Association, 27(8), 1321- bility research on intelligent systems for 1325. doi: 10.1093/jamia/ocaa105 people who are deaf or hard of hear- Page, L. C., & Gehlbach, H. (2017). How ing. ACM SIGACCESS Accessibility an artificially intelligent virtual assistant and Computing, 125, article 4. doi: helps students navigate the road to col- 10.1145/3386296.3386300 lege. AERA Open, 3(4), 1-12. doi: Kapoor, S. K. (2020). New face of indian 10.1177/2332858417749220 governance sector with artificial intelli- Pantea, M. (2019). Limited abilities of gence. IME Journal, 14(1), 97-110. doi: computers as translation devices. In 10.5958/2582-1245.2020.00014.7 T. Mateoc (Ed.), Cultural Texts and Kushalnagar, R. (2019). A class- Contexts in the English Speaking World room accessibility analysis app for - VI - New Perspectives (pp. 228– deaf students. In The 21st In- 233). British Association for Modernist

26 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020

Studies. Retrieved from http:// tion and communication technologies. In culturaltextsandcontexts.ro/ S. E. Burgstahler (Ed.), Universal design proceedings/conference2019 in higher education: From principles to .pdf#page=228 practice (2nd ed.) (p. 275-284). Boston: Popenici, S. A. D., & Kerr, S. (2017). Ex- Harvard Education Press. ploring the impact of artificial intelligence Treviranus, J. (2018). The three dimensions on teaching and learning in higher edu- of inclusive design: A design frame- cation. Research and Practice in Tech- work for a digitally transformed and nology Enhanced Learning, 12(1), arti- complexly connected society (Doctoral cle 22. doi: 10.1186/s41039-017-0062- Thesis, University College Dublin). 8 OCAD University Open Research Rabiner, L., & Juang, B.-H. (2008). Histori- Repository, http://openresearch cal perspective of the field of ASR/NLU. .ocadu.ca/id/eprint/2745/1/ In J. Benesty, M. Sondhi, & Y. Huang TreviranusThesisVolume1%262 v5 (Eds.), Springer Handbook of Speech July%204 2018.pdf. Processing (p. 521-538). Springer. doi: 10.1007/978-3-540-49127-9 26 Trewin, S., Basson, S., Muller, M., Bran- Rohlfing, M. L., Buckley, D. P., Piraquive, J., ham, S., Treviranus, J., Gruen, D., . . . Stepp, C. E., & Tracy, L. F. (2020). Manser, E. (2019). Considerations Hey Siri: How effective are common for AI fairness for people with disabil- voice recognition systems at recogniz- ities. AI Matters, 5(3), 40-63. doi: ing dysphonic voices? The Laryngo- 10.1145/3362077.3362086 scope. Advance online publication. doi: Vaughan, C. E. (1993). The struggle of 10.1002/lary.29082 blind people for self-determination: The Singh, N., & Varshney, U. (2019). Med- dependency-rehabilitation conflict: Em- ication adherence: A method for de- powerment in the blindness community. signing context-aware reminders. In- Charles C Thomas Pub Ltd. ternational Journal of Medical Infor- matics, 132, article 103980. doi: 10.1016/j.ijmedinf.2019.103980 Snyder, T. D., De Brey, C., & Dillow, S. A. (2016). Digest of education statistics 2016 (51st ed.) (NCES 2017-094). Na- tional Center for Education Statistics, Department of Education. Retrieved from https://files.eric.ed.gov/ Natalina Martiniello is fulltext/ED570993.pdf a PhD candidate in the Son, C., Hegde, S., Smith, A., Wang, X., Vision Science (Visual & Sasangohar, F. (2020). Effects of Impairment and Reha- COVID-19 on college students’ mental bilitation) program at the health in the United States: Interview University of Montreal. survey study. Journal of Medical In- Her doctoral research ternet Research, 22(9), e21279. doi: focuses on exploring the 10.2196/21279 factors which influence Story, M. F., Mueller, J. L., & Mace, R. L. the outcomes of working- (1998). The universal design file: De- age and older adults with signing for people of all ages and abili- visual impairments who learn braille, including ties. North Carolina State University, The the role of refreshable braille technologies Center for Universal Design. Retrieved within this context. Her research interests from https://files.eric.ed.gov/ straddle the fields of education and reha- fulltext/ED460554.pdf bilitation, with an undergraduate degree in Thomson, R., Fichten, C. S., Havel, A., Budd, English and Educational Studies from McGill J., & Asuncion, J. (2015). Blending uni- University and a M.Sc. in Vision Rehabilitation versal design, e-learning, and informa- from the University of Montreal.

27 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020

Jennison Asuncion Alice Havel completed a is co-director of the Ph.D. in Counseling Psy- Adaptech Research Net- chology at McGill Univer- work and a member of sity. Before her retirement the advisory council for she was the Coordinator CSUN (California State of the Student Access- University Asesistive Ability Centre at Dawson Technology Conference). He is intimately College. She is a re- familiar with the Canadian and American search associate with the postsecondary education ICT accessibility Adaptech Research Network and a Scholar in landscape. Now residing in the USA, he has Residence at Dawson College. Her research lived in both Toronto and Montreal, where he focus is on the development of inclusive teach- worked on digital accessibility in the ﬁnancial ing practices through universal design and the services industry and volunteered with several use and accessibility of information and com- non-proﬁt organizations. In 2012, he received munication technologies in postsecondary ed- a Queen Elizabeth II Diamond Jubilee Medal ucation for students with disabilities. recognizing his accessibility contributions in Canada.

Catherine S. Fichten, Ph.D. is a professor in the Maegan Harvison is a Department of Psychol- third-year psychology stu- ogy at Dawson College dent at Concordia Univer- and an Associate Pro- sity. She has also been fessor in the Department a research assistant with of Psychiatry of McGill the Adaptech Research University. She co-directs Network for over three the Adaptech Research years where she offers a Network. Her research unique student perspec- interests include factors tive. She hopes to continue gaining research affecting the success of experience and broadening her interests. college and university students with various disabilities, with an emphasis on information and communication technologies.

Mary Jorgensen is currently enrolled in the Anick Legault obtained Master’s for Counselling her Ph.D. in Psychol- at Athabasca University. ogy from the Universite´ She received her BA du Quebec´ a` Montreal.´ degree in Psychology She has been working from Bishop’s University. as a Psychology teacher While at Bishop’s she was at Dawson College since inducted as a Golden Key 2010. She has done ed- Scholar. Her experience includes working as ucational research both in a research and teaching assistant at Bishop’s Canada and in the United University. She is currently a research asso- States. While her doctoral ciate at the Adaptech Research Network. Her research centered on effective teaching meth- research interests include the experiences of ods, her true passion is teaching using ICTs in postsecondary students with mental health an inclusive way, and more particularly using related disabilities and the employment of smartphones in the classroom as a pedagogi- students with disabilities. cal tool.

28 AI MATTERS, VOLUME 6, ISSUE 3 DECEMBER 2020

Alex Lussier completed her undergraduate degree in International Studies in 2019 at the University of Montreal. She has been a research assistant at the Adaptech Reseach Network and Cegep´ Andre-Laurendeau since January 2015. Her research interests include the integration of universal design in postsecondary pedagogy as well as the digitalization of teaching and the new pedagogical approaches that this entails. Christine Vo is a student at Dawson College and a technology specialist for Adaptech.