How to Use LEIA

LEIA (Learning with an Educational Intelligent Assistant) is a platform that enables instructors to design interactive simulations and students to practice in realistic scenarios with AI support.

---

1. General Workflow

The general workflow of LEIA follows four main steps:

1. Instructor designs the LEIA

   • Chooses a process/behaviour (currently, either a conversational game, information requirements elicitacion, or process requirements elicitation).
   • Defines the problem (including the expected solution).
   • Chooses or customizes the persona whith which students will interact. Including personality traits.

2. Instructor configures an experiment

   • Defines the scenario (conditioning of LEIA).
   • Adjusts operational parameters (time limits, replicable sessions, etc.).
   • (Optional) Provides an auto-grader for automated evaluation.

3. Students access the activity

   • Each student/group receives their own dedicated session.
   • The instructor specifies if the activity must be completed in a single session or across multiple ones.

4. Submission and evaluation

   • Students upload their work product (e.g., UML diagrams in Mermaid syntax).
   • The work is evaluated either by an auto-grader or manually.

5. Review of results

   • The instructor can download session data (CSV/JSON), including:
     • Full conversation transcripts
     • Student submissions
     • Metadata and timestamps
   • This supports grading, detailed analysis, or research studies.

Workflow Diagram

---

2. Instructor Experience

LEIA creation/design:

Instructors do this task using the LEIA Designer. Using this application instructors can view a listing of existing LEIAS:

From the listing instructors can create new LEIAs using one of the elements in the listing as starting point:

In the LEIA creation interface users can define:

1. Instructors choose from an existing proces, problem and persona.
2. [Optional] Instructors customize the problem or/and the persona.
3. [Optional] Instructors can test de LEIA and interact with the chat to check the behaviour
4. Instructors confirm the status and save the LEIA as private, and can ask admins (by e-mail) to make it public (after a review process).

Experiment configuration

Instructors define:

• Content:

  • Scenario for LEIA (e.g., a simulated client interview, process discovery interview, or conversational game).
  • How students should submit their solutions.
  • Whether to use an auto-grader or manual evaluation.

• Operational parameters:

  • Time limits.
  • Replicable sessions with unique codes.
  • Selection of the LLM provider.

---

Workbench dashboard

The instructor dashboard allows:

• Viewing all instantiated sessions.
• Inspecting a specific session's details:
  • Metadata
  • LEIA configuration
  • Submission and evaluation settings

3. Student Experience

Access

Students log in using:

• Email address, or
• System-provided code (to ensure anonymity if required).

Interaction with LEIA

• Students initiate a chat session with the assistant.
• They can ask clarifying questions, gather requirements, and iteratively refine their understanding.
• If the session is replicable, they can leave and resume it later.

---

Submissions and Feedback

• Students submit their solution (e.g., UML diagram in Mermaid syntax).
• If enabled, the auto-grader provides immediate feedback:
  • Highlights missing elements or errors.
  • Suggests improvements.
  • Allows students to re-enter the session and refine their work.

---

4. Use Cases

• Requirements Elicitation training: simulate client interviews and model requirements with UML.
• Educational simulations: role-play consulting scenarios or analysis meetings.
• Iterative activities: allow multiple sessions with feedback loops.