Plan a Friday
Open the complete schedule, experiment, challenge kickoff, and teacher setup for a specific unit.
Lecturer guide
Evidence-based learning
Responsible collaboration
CAS Human-AI Collaboration
Find teaching plans, participant assignments and course concepts. Across nine challenges, participants test how to predict, steer and check AI behaviour.
The course shifts focus
FROM“What can AI do?”
TO“How do we design human-AI collaboration so that reliable, responsible outcomes emerge?”
Find what you need
Teaching plans and course concepts
Start with a unit to prepare a class or open its participant assignment. Use the concept pages to connect observations across units.
Explore a literacy
Trace what explains and shapes behaviour across Interaction, Mechanism, Data, System, and Human-Context.
Follow the collaboration levels
See how practice develops from asking and instruction toward reflection and orchestration.
Find an AI fundamental
Locate the concepts, demonstrations, caveats, and questions that explain observed results.
Progression
Nine Fridays, one learning arc
Each unit adds a form of control while contributing evidence to the participant’s emerging personal theory of control.
Unit 1From Asking to Structuring: Designing AI InteractionUnit 2From Structuring to Alignment: Clarifying Before GeneratingUnit 3Why Models Behave Differently: Parameters, Variability, and ControlUnit 4From Opinion to Evidence: Grounding AI in Data and SourcesUnit 5From Answers to Processes: Designing Human-AI Co-ConstructionUnit 6Reflection Loops: Critique, Revision, and Better ReasoningUnit 7Leadership, Responsibility, and Legitimation in Human-AI SystemsUnit 8Designing Workflows: From Single Chat to Reliable AI SystemsUnit 9Security and Trust Boundaries in Human-AI Collaboration
From Asking to Structuring: Designing AI Interaction
Participants compare a direct request with a supplied five-phase collaboration process. They learn to shape context, make decisions and judge results through evidence.
From Structuring to Alignment: Clarifying Before Generating
Participants clarify a task before execution and keep accepted decisions in files. They move from a virtual Markdown workspace to local Python analysis, then use verified results to prepare a report across fresh contexts.
Why Models Behave Differently: Parameters, Variability, and Control
Participants hold a task steady, vary one condition and repeat trials. They use the evidence to decide when variation helps, when it creates risk and which checks are needed.
From Opinion to Evidence: Grounding AI in Data and Sources
Participants learn that fluent output is only as defensible as the information made available to the model. They learn to inspect source selection, retrieval, chunking, citation, and claim support rather than treating a grounded answer as reliable merely because it contains references.
From Answers to Processes: Designing Human-AI Co-Construction
Participants turn a single request into an inspectable process with consequential human decisions. Four cumulative exercises use the same equipment-loan case: introduce three process cuts, make the intermediate artefacts and checkpoints explicit, add reviewing perspectives and roles, then evaluate quality, control and effort. PDCA brings these elements together. A live demonstration shows how project folders, Markdown files and skills support continuity and reuse.
Reflection Loops: Critique, Revision, and Better Reasoning
Participants organise critique, evidence and human decisions into workflows. In the Arena they first improve a judge, then keep the selected judge fixed while testing a generator change.
Leadership, Responsibility, and Legitimation in Human-AI Systems
Participants learn that a plausible AI recommendation does not become a legitimate decision merely because a human approves it. They learn to design and test the decision rights, evidence requirements, escalation routes, and opportunities for contestation that make human responsibility substantive rather than symbolic.
Designing Workflows: From Single Chat to Reliable AI Systems
Participants learn to move beyond a sequence of chat turns and design a controllable agentic system with explicit state, capabilities, routing, validation, permissions, stop conditions, and human approval. They learn when orchestration improves reliability and when it only adds complexity.
Security and Trust Boundaries in Human-AI Collaboration
This final Friday makes security concrete as a control problem. Participants integrate learning from the preceding units to recognize where an AI-supported office task can lose control and to redesign the collaboration so that data, instructions, permissions, actions, and human responsibility remain appropriately separated.
Module 2
From accumulated evidence to adaptive mastery
Participants select and interpret evidence from the nine challenges, redesign a context-specific work practice, stress-test its adaptability, and articulate a Human-AI Collaboration Blueprint and personal theory of control.
Open the Module 2 orientationFind guidance
Search the course
Search units, challenges, literacies, levels, fundamentals, teacher setup, and assessment guidance.