Page 62 - DP-2-2

Design+ Da Vinci AI Tutor in art history learning

Figure 6. Chung the Artist’s Palace Louvre Interior-Salon Carré Pack Figure 7. Tiffani Barner’s Da Vinci AI Tutor (2024), developed using
(2021) on Unity Asset Store Convo.ai and Unity

solutions with the specific needs of the project, balancing understanding that the shared storage solution would
functionality, historical accuracy, and accessibility to create remain active for 60 days. Plans for a more permanent
an impactful and immersive educational tool for the final hosting solution were discussed to ensure the continued
version (Figure 7). availability of the application beyond this window.
3.2. Deployment of the Da Vinci AI Tutor to students Technical challenges during this phase initially included
issues with building the application as a web-based solution.
Following successful early testing phases, the tutor Troubleshooting revealed that these errors were linked
was formally deployed across a diverse set of courses to microphone support, a feature integral to the tutor’s
during the Fall 2024 semester. This deployment included voice-to-voice interaction capabilities. Recognizing the
undergraduate and graduate Renaissance Art classes, importance of resolving these issues to enhance platform
global survey courses designed for non-majors, and a compatibility, the development team prioritized adapting
Comprehensive Examination course for graduate students. the build to ensure a smoother user experience. While
The range of courses allowed for the evaluation of the the initial PC version allowed up to 100 interactions per
tutor’s efficacy across varied academic levels and learning day with the Da Vinci avatar through Convo.ai Convai’s
objectives, providing a robust foundation for analyzing infrastructure – a limit that proved sufficient for class sizes
its impact on accessibility, engagement, and learning during the early deployment – feedback from students
outcomes. highlighted significant difficulties with downloading and
The initial iteration of the tutor was available exclusively executing the required installation steps. This complexity
as a PC version, requiring students to download and install created barriers to accessibility, particularly for users
the full Unity build on their desktops with instructions unfamiliar with extracting and launching application files.
provided in Table 3. This version incorporated a newly Technical challenges during the implementation
designed main menu and a basic pause menu for ease of phase included issues related to building and deploying
navigation. The main menu included “Play” and “Quit” the application effectively as a web-based solution. Initial
options, enabling students to seamlessly enter or exit the troubleshooting efforts revealed that errors were primarily
application. Settings were accessible directly within the associated with the integration of microphone support, an
game itself, streamlining the interface for users. Students essential component for enabling seamless voice-to-voice
could exit the game by pressing the “Escape” key and then interactions between students and the Da Vinci avatar.
clicking the “X” at the top-right corner of the application, Resolving these issues was prioritized by the development
though additional features to allow direct exiting through team, given that student interactions heavily relied on
the “Escape” key were under consideration. robust and responsive voice communication. Iterative

The distribution of the build involved sharing a refinements were driven directly by student feedback,
downloadable ZIP file that contained the application. notably influencing improvements to response accuracy,
Clear instructions were provided to ensure accessibility for interaction latency, and synchronization between avatar
all students, including step-by-step guidance for extracting animations and audio output. For example, students
and launching the application. This attention to detail reported noticeable delays in voice responses and
addressed the varying levels of familiarity with software occasional inaccuracies in pronunciation, prompting
installation among the student population. Permissions developers to fine-tune the speech synthesis parameters
were updated to allow easy access to the link, with the and enhance the underlying model training.

Volume 2 Issue 2 (2025) 12 doi: 10.36922/dp.8365