Description
Session Description
A high-quality educational resource entails choice of media that enables the learners to meet the desired learning outcomes (Lane, 2010). But, the current go-to approach when matching learning resources to students with perception related disabilities is following a one-size-fits-all set of standards (ISO/IEC Information Technology Task Force (ITTF), 2012) in the form of pre-determined filters based on specific content features. Given that there are different drivers of quality that are present in educational resources, this is not ideal. Furthermore, the quality of learning experience relies greatly on the individual’s self-assessment of what filters apply to them. This can be a sensitive question especially when it comes to people with disabilities. Artificial Intelligence (AI) and personalization give us the opportunity to break away from this frame and match diverse Open Educational Resources (OERs) to diverse learners through understanding and calibrating materials to personal abilities and preferences of individuals. Prior work in applying AI in Education has led to developing models that can constantly learn from learner actions and adapt to the learner’s preferences (Sahan Bulathwela, 2020). Through this work, we aim to develop an intelligent educational recommender for learners with visual impairment. Our system is inspired by (Sahan Bulathwela, 2020), a scalable probabilistic graphical model that is rich in transparency. This model can start with basic assumptions (or no prior assumptions) about the learner and rapidly learns a fully personalized model for everyone. The features that the model learns about the student are based on ISO accessibility standards (ISO/IEC Information Technology Task Force (ITTF), 2012). Because the model can start without prior information about the student’s disability, the model will quickly learn the student’s preferences based on their interaction with OERs. This also means that defining prior assessments about the learner disabilities is not essential for the functioning of the system. Having transparency by design, the learned parameters for individuals can also empower the teachers to recognize individuality of different learners and provide them with better support. The developed system is currently being tested with high school students with visual impairments and their teachers using pilot studies. We aim to understand how this personalization model can match learners to OERs while giving better visibility to their teachers to better support them.
Conference presentation will focus on model description (as it is a novel ML approach), recommendation engine and teacher portal service demonstration, and presentation of results from pilot study about service usability with blind high school students.
The research described in this paper was conducted as part of the X5GON project (www.x5gon.org) which has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement
No 761758
References
ISO/IEC Information Technology Task Force (ITTF). (2012). ISO/IEC 40500:2012 W3C Web Content Accessibility Guidelines. Technical Committee : ISO/IEC JTC 1 Information technology.
Lane, A. (2010). Open Information, Open Content, Open Source. In The Tower and The Cloud (pp. 158-168).
Sahan Bulathwela, M. P.-O.-T. (2020). TrueLearn: A Family of Bayesian Algorithms to Match Lifelong Learners to Open Educational Resources. AAAI Conference on Artificial Intelligence 2020.
-
Teresa MacKinnon joined the session Accessibility recommendation system [O-113] 4 years, 7 months ago
-
-
Anna Page joined the session Accessibility recommendation system [O-113] 4 years, 7 months ago
-
Stacy Katz joined the session Accessibility recommendation system [O-113] 4 years, 7 months ago
-
-
joined the session Accessibility recommendation system [O-113] 4 years, 7 months ago
-
Leo Havemann joined the session Accessibility recommendation system [O-113] 4 years, 8 months ago