Page 57 - DP-2-2

Design+ Da Vinci AI Tutor in art history learning

3. Methodology sessions, the appropriate demeanor and accuracy were
attained before the 3D animation build. The overview of
3.1. Avatar design and production phases training steps and prompting can be found in Table 1.
The development of the Da Vinci AI followed a Once the initial training phase was completed, the
structured series of design stages, each marked by distinct model was integrated into the Da Vinci avatar. This
considerations, challenges, and critical decisions (Figure 1). required additional calibration to synchronize the
These stages ensured that the tutor was accessible, linguistic capabilities of the model with the interactive
engaging, and effective for a diverse range of art history elements of the avatar. Multiple rounds of iterative testing
students. The initial stage focused on conceptualizing the followed, where the avatar was exposed to simulated
design and functionality of the tutor. At the heart of this tutoring sessions. During these sessions, developers posed
phase was the decision to use Unity as the game engine, a range of questions and scenarios, evaluating the ability of
chosen for its compatibility with existing plugins such as the model to respond accurately, maintain conversational
OpenAI, as well as its support for features such as rigged coherence, and exhibit personality traits consistent with
avatars, speech synthesis, and integration with language the art historical information desired, as seen in Table 2.
models. The avatar design prioritized historical accuracy Feedback from these sessions informed subsequent
while maintaining an approachable demeanor, embodying refinements, focusing on areas such as response accuracy,
Leonardo da Vinci as both an authoritative and accessible conversational flow, and the ability of the avatar to adapt to
figure. As such, initial design meetings included designing diverse student queries.
a personality, as well as a voice and style guide for the Once the model reached a stage of functional maturity,
avatar. Through a series of iterative prompt engineering the next phase involved integrating the model into Unity,

Figure 1. Infographic of the research phases for Da Vinci AI Tutor
Abbreviation: AI: Artificial intelligence.

Volume 2 Issue 2 (2025) 7 doi: 10.36922/dp.8365