Abstract:

Artificial intelligence (AI) offers novel solutions to the Two Sigma Problem, highlighting the superiority of one-on-one tutoring over traditional classroom instruction. The SMARTA project, leverages a chatbot system to narrow this educational achievement gap through personalized AI assistance.

Featuring three specialized chatbots integrated into a campus management system, SMARTA aims to enhance student motivation, engagement and outcomes. Since 2023, over 5000 students at a German university have access to these chatbots, benefiting from a system that tries to mirror the personalized and adaptive nature of one-on-one tutoring. SMARTA endeavors to evaluate the efficacy of AI chatbots in mimicking the depth of personal tutor-student interactions, thereby potentially reducing the performance disparity among students.

The three chatbots have different tasks and personalities:
a) ALIX, the first of the chatbots, addresses the motivational aspects of student life by offering personalized interactions that are designed to simulate empathetic support effectively, aiming to improve well-being and mental health in a manner that students perceive as genuine, thereby supporting a foundational element of academic success.
b) ROBYN seeks to deepen students' engagement with their studies by encouraging personal reflection on specific academic topics through dialogue-methods such as the Socratic questioning technique, thereby fostering a self-directed learning approach reminiscent of one-on-one tutoring.
c) MELLY aims to actively engage students in a learning dialogue by initiating conversations and leading the discussion, using content that is aligned with the curriculum, thereby simulating the adaptive and responsive nature of personal tutoring.

Integration and Personalization:
The chatbots’ seamless integration into the campus management system enables the use of existing student data such as age, gender, major, current study load, and academic performance to personalize initial interactions, thereby replicating the informed, adaptive responses characteristic of personal tutors. This approach facilitates a level of personalization and relevance in conversations that can enhance learning outcomes.

Monitoring:
A monitoring framework tracks engagement without recording specific conversation content, focusing instead on metrics like conversation length, sentiment changes, and overall engagement. This data supports the ongoing assessment of the chatbots’ effectiveness in mimicking the beneficial aspects of one-on-one tutoring.

Future Developments:
Future enhancements will enable the chatbots to reference past interactions, allowing them to increasingly understand and adapt to the individual needs and preferences of their students over time. By further personalizing the learning experience through building upon previous dialogues, this continuous interaction is designed to mirror the cumulative and adaptive nature of personal tutoring, deepening the chatbots' comprehension of each student's unique academic journey.

Conclusion:
By leveraging AI to provide personalized, interactive support, the SMARTA project represents a forward-thinking approach to addressing the Two Sigma Problem in higher education. Through ongoing development and empirical evaluation, the project aims to demonstrate the potential of AI-assisted learning to narrow the educational achievement gap, turning the Two Sigma problem into a Two Sigma opportunity.

Keywords:

Artificial intelligence, Two Sigma Problem, one-on-one tutoring, SMARTA project, AI assistance, student motivation, Socratic questioning, campus management system, TraiNex, sentiment analysis.