Inspiration
Imagine you’re a student facing a digital exam. Instead of relying solely on your knowledge, you habitually and mindlessly resort to using ChatGPT and Google for answers. Currently, nearly 90% of students admit to using ChatGPT for their academics, a detrimental practice to both memory and critical-thinking skills. This alarming trend motivated us to tackle this issue head-on by harnessing voice-to-text technology to enhance information retention. By encouraging students to verbalize their answers, we aim to improve communication skills and cultivate quick thinking, akin to the interactive learning experience of raising one's hand in a traditional classroom, now replicated in a virtual setting.
As our idea evolved, we realized that this software could have further use in the hiring process. Currently, the hiring manager platform Hirevue has completed 70 million asynchronous online interviews. Our AI-powered plagiarism detection system safeguards against scripted responses, fostering authenticity among applicants. We believe that academic and workplace integrity is paramount, which motivated us to create EvaluAI.
What it does
Think of us as GPTZero for speech submissions.
Currently, we have two use cases pending integration with preexisting services: a Learning Management System (e.g. Canvas, Blackboard, Google Classroom, etc.) and asynchronous virtual hiring platforms (e.g. HireVue).
On the Learning Management System (LMS) interface, professors access our web app dashboard to upload lecture slides or textbook material for generating quiz questions. Our platform then automatically creates a pre-quiz based on this content and distributes it to student dashboards. Upon logging in, students select the quiz and are presented with a free-response question. They have a set time limit before they must begin verbally responding via webcam. Leveraging AI generation and plagiarism detection APIs, we ensure students craft original answers, deterring reliance on scripts from ChatGPT or similar tools. Results, including AI detection and plagiarism percentages, are showcased on the professor dashboard, with successful completion granting access to the actual quiz on the LMS (upon integration). This additional layer of security reinforces the necessity for students to prepare actively rather than resorting to rote memorization.
Similarly, in the interview segment, Hiring Managers provide job descriptions, prompting our platform to generate a specified number of interview questions. This approach discourages candidates from simply copying their resumes and using ChatGPT to script responses.
How we built it
Our initial priority was to incorporate a reliable speech-to-text algorithm, recognizing its pivotal role in our product's functionality. Once this cornerstone was established, we divided our team to focus on either frontend or backend tasks, based on individual expertise.
Initially, backend efforts were directed towards implementing Auth0 for authentication and integrating the GPTZero API to identify AI-generated content in the text version of responses. Meanwhile, frontend developers concentrated on integrating webcam functionality using React.
Challenges we ran into
We encountered challenges while integrating various APIs, including OpenAI and GPTZero, as well as during the implementation of the speech-to-text conversion process. The conversions frequently proved to be inaccurate and occasionally failed altogether.
Accomplishments that we're proud of
We take pride in bringing our vision to fruition through the integration of various cutting-edge technologies. Leveraging the OpenAI, GPTZero, and AssemblyAI APIs, we built a robust backend using Flask, while our frontend was crafted using React.js, HTML, JavaScript, and CSS. Authentication was streamlined with Auth0, and data was efficiently managed using MongoDB.
Our motivation stems from the tangible impact our project can have on promoting academic integrity, enhancing students' memory retrieval capabilities, and reducing reliance on tools like ChatGPT and Google for open-book exams. With a staggering 89% of students admitting to using ChatGPT for homework and 84% of teachers expressing concerns about its potential for facilitating cheating, the urgency for such technology is undeniable.
We're proud to have taken the initial steps toward realizing our goal of safeguarding human intelligence in an era of rapid AI advancement.
What we learned
Through this project, we gained valuable insights into managing multiple APIs and leveraging third-party services effectively. Additionally, we acquired expertise in seamlessly integrating frontend and backend components to create a cohesive and functional product.
Furthermore, we acquired the knowledge and skills required to integrate AI seamlessly into our web application, enhancing its functionality and utility. These experiences have not only enriched our technical abilities but have also equipped us with valuable tools for future projects in the ever-evolving landscape of technology.
What's next for EvaluAI
Our future plans involve further development of the frontend, with a focus on integrating a feedback feature following student responses. Additionally, we aim to refine the application to cater to the needs of teachers, school administrators, hiring managers, and students or job applicants alike.
We are committed to bringing this technology to higher learning institutions and companies that utilize virtual asynchronous interviews in their hiring processes. By bridging the gap between academia and the workforce, EvaluAI will provide a comprehensive AI-driven solution for education and the interview process, serving as a valuable tool for both students and professionals alike.
Log in or sign up for Devpost to join the conversation.