Trust and ethical considerations in a multi-modal, explainable AI-driven chatbot tutoring system: The case of collaboratively solving Rubik's Cube

TL;DR

This paper addresses trust and ethical considerations in a multimodal, explainable AI tutoring system (ALLURE) designed to collaboratively solve Rubik's Cube with high school students. It combines DeepCubeA-based reinforcement learning with inductive logic programming to produce explainable macro actions, which are then translated into student-friendly natural language explanations via NLG; additional components for sentiment analysis and privacy protection are integrated to ensure safe interactions. The authors highlight two core trust challenges—ensuring acceptable conversations and preventing information leakage—and present a structured evaluation plan, including automated tests, user studies, and security assessments, to validate the system in educational settings. By providing an explainable, privacy-preserving tutoring interface, the work advances AI-assisted education and sets the stage for future field deployments, while grounding bias and safety considerations in formal mechanisms such as backdoor-adjusted causal analysis $P[Y|do(X)]=\sum_Z P(Y|X,Z)P(Z)$ and related DIE/WRS metrics.

Abstract

Artificial intelligence (AI) has the potential to transform education with its power of uncovering insights from massive data about student learning patterns. However, ethical and trustworthy concerns of AI have been raised but are unsolved. Prominent ethical issues in high school AI education include data privacy, information leakage, abusive language, and fairness. This paper describes technological components that were built to address ethical and trustworthy concerns in a multi-modal collaborative platform (called ALLURE chatbot) for high school students to collaborate with AI to solve the Rubik's cube. In data privacy, we want to ensure that the informed consent of children, parents, and teachers, is at the center of any data that is managed. Since children are involved, language, whether textual, audio, or visual, is acceptable both from users and AI and the system can steer interaction away from dangerous situations. In information management, we also want to ensure that the system, while learning to improve over time, does not leak information about users from one group to another.

Figures (10)

• Figure 1: The information that facilitates collaboration between human and AI that needs to be protected from leakage.
• Figure 2: The Solved white cross
• Figure 3: Architecture of a general data-driven collaborative assistant and our specialized components.
• Figure 4: A working screenshot of the conversation illustrated in the Table \ref{['Tab:CD']}
• Figure 5: A macro action used for solving the White Cross. (a) Preconditions: An edge piece has a white sticker on a face that has the same color as its other sticker. The other sticker is facing the yellow face. In this case, the other sticker is orange. (b,c,d) Sequence of actions: D', F', R, F (e) Effect: The edge piece is now in place.