Smart Quiz+

Children with special needs are always let down during exams, and almost never because teachers don’t care about their welfare. It is up to the teacher to manually insert extra time, font size, and the Text to Speech options for each student for each exam. Proctoring software identifies screen readers and any other assistive technology as cheating tools. The language in the exams is also very complex that it fails children with dyslexia, ADHD, or language processing disorders in the language and not in the subject.

This is why we decided to develop a tool that is free and considers the needs of the child. What it does

SmartQuiz+ is a free and accessible exam platform. It already includes:

Voice controls: students can navigate the exam through voice commands. Text-to-speech: questions and instructions can be read aloud. Speech-to-text: students can dictate their answer.

We have also developed and prototyped the next layer of functionalities:

Accessibility profiles: a student configures his or her accommodations once – such as extra time, font, contrast, text-to-speech – and it will be automatically applied to all exams with live preview. Plain language version of the text: a teacher hits “Simplify” button on the question, gets the original text and a more understandable suggestion. Teacher decides to approve, modify or reject this suggestion, and the original text remains. Simplified version is visible only for those students who need such accommodation. Assistive technology integrity dashboard: declared assistive technologies are never considered to be suspicious. All flags go to human review without any automatic penalties. Accessibility metrics and regional language support: teachers receive information about which questions were harder for assistive technology users, and the platform interface is available in English, Tamil and Hindi languages, as well as offline connectivity indicator.Construction process

SmartQuiz+ was developed by Lovable using Vite, React, TypeScript, shadcn-ui, and Tailwind CSS. For voice features, we use either Web Speech API [or please name what was used by Lovable]. We selected shadcn-ui because its components are based on accessible primitives, providing us keyboard navigation and ARIA support out of the box; we then improved it by adding labels, focus states, and contrast tests.

To develop the new features, we first implemented the UI part, testing the workflow with teachers and students while still working with mock data. The simplification through AI is to be applied once the teacher is creating the exam; it will be saved and won’t be performed again during the test; thus, it has no cost for each student and cannot be used to cheat. AI is not going to see any data but the question text. Challenges we encountered Avoiding oversimplification by making the questions simpler. Making something simpler may cause changes in the question which might make it easier or change its meaning. Our solution is a side-by-side comparison where everything needs approval by the teacher, along with the ability to disable simplification of the questions used to measure linguistic skills. Excluding AI from the testing process. From day one, we decided that the use of AI would be to assist the preparation of the content but never be helpful to the learner in answering the questions. Designing for accessibility, not just accessibility. Navigation by keyboard, labeling for screen readers, and proper contrast were things that needed attention on the very first screen. Voice options across different devices. Voice recognition works differently in different browsers, accents, and languages, therefore, typing was always an option. Sensitive Data. Information regarding the disability is personal. The exam receives only the required information, like “extra time 25%.”What have we done that we’re proud of? An accessible test experience through voice, text-to-speech, and speech-to-text capabilities. The design of AI that by default is secure: teacher approved, authoring time only, and always keep original version. A prototype of the whole workflow of both teacher and students, including profile creation to simplify and integrity check. To remain the platform open-source so even a small school or a teacher on its own can use it.Learnings we have gained Accessibility should be designed at the initial phase rather than being added later. AI in education is of more use when it aids the teacher and does not interfere in the thought process of the student. The reading level and the level of subject knowledge are distinct concepts, which examinations tend to mix. First prototyping of the interface proved quite effective for testing ideas before developing the other aspects.Future directions for Smart Quiz+ Integrate the Simplify module with an actual AI engine while ensuring teacher approval and using the initial text. Save access profiles safely to make sure they transfer with the student between tests. Pilot with one or two schools using NVDA, TalkBack, and VoiceOver. Include real Tamil and Hindi support in the interface and the simplification module, after checking with native speakers. Create offline functionality that makes it possible to take tests with bad internet connection and synchronize later. Make the app open-source with funding from grants and paid hosting services for big organizations.