SDG 4: Quality Education
SDG 10: Reduced Inequality
2. Project Details
Company or Institution
University of Tuscia – DEIM
General description of the AI solution
The AI solution we propose, named BESPECIAL, is a software platform able to provide dyslexic students at university with customized digital supporting tools and personal learning strategies. The kernel of the platform consists in an AI-based module that is capable of predicting the specific support needed by each user, from his/her dyslexia clinical report, the answers to a self-evaluation questionnaire about the difficulties faced during the studies and the solutions they deem to be helpful, and a battery of psychometric tests. AI is trained on a large database of clinical reports of dyslexic students and on questionnaires and psychometric tests performed by them. The output of BESPECIAL are digital tools (e.g. audiobook, concept maps etc.) and best practices (e.g. presence of tutors, different examination procedures etc.) that proved to be more useful for each student. Digital tools will improve the students’ study material and adapting it, using AI, based on students’ needs, whereas the services will be provided to the higher institutions, in order to define standard strategies for inclusive education. In fact, well-known techniques like NLP, OCR, GANs etc. are employed to adapt the material to the specific needs of each student. The above-described AI solution has been conceived within a European project, named VRAILEXIA (www.vrailexia.eu), which aims at overcoming all the main difficulties encountered by dyslexic student at university and reducing the gap with respect to non-dyslexic students. In addition to BESPECIAL development, VRAILEXIA will lead to: (i) implement a battery of virtual reality test to collect in real-time the skills score of dyslexic students; (ii) realize of an online shared repository; (iii) create a training path for dyslexia awareness and (v) a memorandum of understanding to spread common inclusion strategies among higher education institutions.
University of Tuscia – DEIM
Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.
The AI application we propose has a high degree of novelty. Indeed, AI has been applied in previous works aimed at diagnosing dyslexia, but no relevant AI solutions have been proposed yet to support dyslexic students during their academic career.
One of the main issues is the heterogeneity of dyslexia manifestations. To deal with it, the AI assessment module underlying our application is fed with different kinds of input, namely clinical diagnosis, students’ self-evaluations and objective test scores, which, together, embrace the entire spectre of dyslexia-related problematics. The AI module is divided into two parts. The first one provides general supporting tools and strategies for different categories of dyslexic students, which are also determined by the own AI module. It consists of an innovative machine learning algorithm that combines the potential of data- and human-driven approaches. This allows mixing the information coming from the collected data with the corpus of knowledge about dyslexia, so as to improve the training phase and provide predictions that are both accurate and meaningful. The second part of the AI module, instead, makes the supporting tools and strategies specific for each single student. To do this, dyslexic students are monitored while using them and specific indicators of their performances are collected and feed back to AI. This, in turn, upgrades its predictions toward a customized support offer. Big data techniques are employed, in order to be capable to treat the large amount of information coming from the performance indicators. At the time of submission, BESPECIAL is related to TRL3 – Proof of concept, as also demonstrated by paper (DOI: https://doi.org/10.3390/app11104624) aiming to reach TRL5/6 within the end of the year. BESPECIAL was presented in social media and through media events, especially in Italy. It will be also presented in international conferences in the next months.
Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.
BESPECIAL meets the SDG 4 and 10, specifically the targets 4.5 and 10.2, since it aims to a more inclusive and equitable quality education counteracting the dropout rate increasing among SLD students. It will customize itself and adapt its output contents to the dyslexic students’ needs based to foster all students to enhance their strengths and untap their potential.
The effectiveness of the AI solution will be monitored; BESPECIAL will be tested quarterly with three data collection phases within the international partnership. BESPECIAL is characterized by scalability and transferability that will permit to spread it in different languages and teaching areas. BESPECIAL can be easily scaled and consequently its impact to SDG for any other learning disorders, transferred and tailored for the working environment. The impact to SDGs will be measured considering the : (i) reduction of dropout rate; (ii) increasing of dyslexic students enrolled at Universities; and (iii) reducing of the ECTS gap obtained by SLD students.
Beyond the project, BESPECIAL will guarantee the use of AI as technical solution for ensure a more inclusive and quality education within Higher Education leading the development of an inclusive ecosystem within HEIs. Such solution has the ability to ensure a wide and balanced impact across a high number of UNESCO Member States. In fact, the solution is the result of a collaboration among European states (Italy, Spain, France, Belgium, Greece and Portugal) and it will be provided at the end of the project in all the national languages and in English. This guarantees the possibility to spread the solution across several UNESCO member. The AI solution will be also the provider for the correct guidelines to follow for the realization of the above-mentioned memorandum of understanding, which leads to the development of an inclusive ecosystem of higher institutions.
Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.
Several target groups can be identified for the evaluation of the impact of the proposal. Dyslexic students, as primary end users, will be able to use an innovative tool and become part of a self-evolving virtual creative ecosystem. The main impact will be related to the enhancement of the motivation in learning process and to the introduction of equal opportunities for dyslexic students to actively participate in HE. As a more general impact, a stronger social and cultural inclusion will be guaranteed not only in the university life but also in labour market. Positive impacts can be associated also to teachers, since they will be able to understand more deeply the problems induced by dyslexia. Finally, university as a whole will increase the quality of students’ management both in the enrolment phase and after, by supporting them after the Degree achievement. The scalability of the project is guaranteed by the possibility to: (i) access the platform (it will be hosted in an open source repository) and the related documentation, which will allow anyone to update it with new possible AI solutions; (ii) extend the application to different targets of students, as those affected by other learning disorders. The proposed solution could be exploited by other research groups or also companies that, working on AI, can focus their production line on the realization of tools for the digitalization of the study materials based on the outputs suggested by BESPECIAL. Although at a first view such solution can be thought for a restrict group, it should be considered that, according to recent statistics, the number of dyslexic students attending university is constantly increasing with an exponential growth. As an example, we collect in less than three months more than 1300 answers to the questionnaire only considering Italian students.
Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.
Ethical challenges posed by artificial intelligence are related to its powerful applications in the scientific research and clinical practice of healthcare. However, unless these challenges are tackled at their core, they will likely lead to discrimination, including, but not limited to, disability, gender, ethnicity, and age. To avoid making unfair, biased, and unethical decisions that might reflect wider prejudices in society, data sets that are used to train AI algorithms should be highly representative of the wider population, for example, the work plan ensures that students with specific learning difficulty such as dyslexia are highly represented regardless their ethnic group, gender, age, geographic regions, and social realms. Equally important, AI systems do not discriminate between the range of abilities (e.g., population at the severe end of the dyslexia continuum), social class, cultures, and economic backgrounds. We encourage AI solutions for dyslexic healthcare students that are evenly distributed, and thus avoiding negative effects on individuals with rare medical conditions, or others who are underrepresented in research data. Trust and access to patient data also raise important ethical issues that will likely result in trust crisis about the AI technology being developed under the pretext of public interest. Practically speaking, in this context, dyslexic students, researchers, partners, and relevant stakeholders will need to be able to trust AI systems, so that these systems can be implemented successfully in dyslexic healthcare. As AI has the potential to be used not only to ‘do good’ but also for malicious purposes that violate the bioethical principle of ‘do no harm’, another pressing question that has received enough attention in the project concerns participants data; initiatives of using data that raise privacy concerns go beyond the legal compliance to take account of dyslexic students’ expectations about how and for what purpose their data will be used.