The presence of generative artificial intelligence (gen AI) and Large Language Models (LLMs) has become ubiquitous. Students are already using gen AI, often to help them understand complex ideas, brainstorm assignments, or check their writing.
Introducing a Cogniti agent into your unit allows you to bring that activity into your learning design, ensuring students engage with an AI tool that reflects your disciplinary language, expectations, ethics, or perhaps a structured way to learn.
Creating an agent is only the first step. Like implementing any learning design tool or strategy, Cogniti agents benefit from ongoing review and iteration. The most effective agents evolve alongside their students, shaped by feedback, usage insights, and educator reflection, and are constructively aligned to the learning outcomes.
1. Introduce Cogniti in the context of the learning outcomes
Start by introducing your Cogniti agent directly in class or online (on Canvas). Even though students might be familiar with Gen AI, many students may not have heard about Cogniti and why they are being asked to use a different tool. Show students how the agent you have designed connects to your unit’s learning outcomes and how it can support their learning goals. A good strategy is to follow four steps for implementing custom AIs that help students learn, not outsource:
- Give students the context
- Model how to interact with Cogniti agents
- Emphasise the value of learning
- Invite reflection and feedback
Why do we do this?
When you explain why the agent exists, for example, to help them unpack complex concepts, practice academic writing, or explore case studies, you support students’ intrinsic motivation. According to Self-Determination Theory (Ryan & Deci, 2000), students are most engaged when they experience autonomy, competence, and relatedness. A well-framed Cogniti agent can nurture all three: students can explore independently, receive guidance at their level, and feel that the tool is designed for them within their course.
It also helps to clarify how Cogniti differs from general tools like ChatGPT. While public models draw on broader contexts, Cogniti agents can be specifically contextualised: built around your learning outcomes, the student’s individual learning activity and assessments, and the University’s wellbeing and integrity frameworks.
2. Use Cogniti insights and conversation history to guide improvement
Cogniti includes Insights and Conversation History, dashboards that reveal how students are interacting with your agent. Reviewing these can help you identify patterns to refine the system message.
For instance, in BUSS1000 during Semester 2, the teaching team observed that 12% of students sought advice from a Presentation Mentor agent regarding their well-being. Upon reviewing feedback for the agent, the teaching team adjusted the system message to include: “If the question provided by a student is of a student wellbeing issue, please answer in a sensitive manner and redirect the student to [Student Wellbeing Services](https://www.sydney.edu.au/students/health-wellbeing.html)”. This small, data-informed change ensured that the agent remained helpful yet supportive, and without overstepping its purpose in its focus to support students with developing and practising their presentation of a business pitch.
This process mirrors design-based research approaches like the ones reported in Wang & Hannafin (2005), where interventions are refined through cycles of implementation, reflection, and redesign informed by learner data.
To access insights and conversation history, find these in the Cogniti Dashboard, with the ‘view insights’ and ‘view conversation history’ icons:
3. Ask your students what they think
While analytics show what students do, direct feedback reveals what students may feel about the use of customised AI agents. You can embed a quick Student Relationship Engagement System (SRES) survey as seen in this example or contact students directly to ask:
- What do you like about the customised AI?
- How could we improve on using customised AI?
- Any further comments?
Perhaps surprisingly, student surveys have revealed nuanced perceptions of the trustworthiness and value of AI tools (Ardelean and Edit, 2023), indicating that student perceptions of technology often influence how they utilise these tools. Inviting and being responsive to learner feedback when introducing learning tools provides a clear signal to students that they are partners in an evolving learning experience, rather than passive recipients of technology.
You don’t need to wait for perfect data! Even a handful of comments from class discussion can inspire valuable tweaks to tone, clarity, or guidance.
