AI assistant MVP
An AI assistant could bring multiple benefits for Moodle users.
- We want users to interact with Moodle through user friendly interfaces. An AI assistant could be a new user interface that would grant quick access to the most popular user stories different stakeholders have in Moodle.
- We want Moodle to actively nudge users about different aspects of their interest. We currently use the notifications system in Moodle to display insights from the analytics API. An assistant could better channel these insights to the user.
- An AI assistant could monitor the user session and actively help or encourage the user when necessary. There are some examples of this in this proof of concept: https://github.com/dmonllao/moodle-local_amigo
- An AI assistant should understand users' input in natural language (NLU) and could respond to them, forwarding them to external documentation pages or asking further questions about their intent. There is a proof-of-concept of Moodle connecting to a NLU backend in https://github.com/dmonllao/moodle/tree/ai-assistant. The responsibility of the NLU backend in this POC is to identify the user intent and/or to reply to user questions.
As an open source learning platform whatever solution we design should be highly customisable and extendable so we should abstract parts of the AI assistant API so we allow users to use different NLU backends and we allow moodle components (including 3rd party plugins) to hook into the assistant.
Some sources of information about how an AI assistants (or chatbots) can help users in an LMS:
- https://github.com/microsoft/Moodle-Teams-Bot (I haven't been able to locate the list of intents yet)
The first thing we should do is to identify how an AI assistant adds extra value for Moodle users. All ideas are welcomed.