CAS Human-AI Collaboration

Module 2

The thesis-like integrative project that turns Module 1 evidence into a Human-AI Collaboration Blueprint and personal theory of control.

Go to project focusDownload Module 2 PDF

Summary#

Module 2 is the CAS integrative project. Participants use evidence from Module 1 to redesign a professional, disciplinary or organisational work practice.

The result is a Human-AI Collaboration Blueprint and a personal theory of control: an evidence-based explanation of what makes collaboration work in this context, where its limits lie and how it should adapt.

The guiding question is:

How can reliable, responsible collaboration be maintained when tools, models, data, policies or working conditions change?

Position in the CAS#

Module 1 develops interaction, alignment, grounding, co-construction, reflection, responsibility, orchestration and security practices. Its experiments also build an understanding of model behaviour.

Select the portfolio evidence relevant to the project. Connect and critically interpret it, including failed assumptions and unsuccessful interventions. There is no need to reproduce every challenge or use every artefact equally.

Module 2 extends these bounded experiments into a sustained inquiry in one context.

Project Focus#

Choose a task, document, decision process or workflow from your own context. Diagnose the current practice and design a more controlled, resilient and responsible way of working with AI.

Use the five literacies where they matter:

  • Interaction: requests, clarification, examples and feedback.
  • Mechanism: variability, context limits and generation conditions.
  • Data: sources, grounding, retrieval and traceability.
  • System: workflow stages, roles, tools, checkpoints and handoffs.
  • Human-Context: trust, responsibility, legitimacy, disclosure and accountability.

Identify primary and supporting literacies. Explain where a literacy is less relevant. They should contribute to one coherent design.

Choose collaboration modes deliberately: instruction, alignment, co-construction, reflection, orchestration or a combination. A higher level is not automatically better.

Demonstrated Mastery#

Participants demonstrate adaptive mastery when they can:

  • diagnose weak, implicit or fragile control;
  • predict how a change should help and design a testable intervention;
  • produce inspectable work and credible evidence;
  • interpret results through relevant literacies and collaboration modes;
  • distinguish evidence from fluency, confidence, convenience or stakeholder acceptance;
  • define human decisions, safeguards and adaptation rules;
  • explain limits, residual risk and transfer to another context.

Reflection must be supported by evidence.

Suggested Project Logic#

Task#

Redesign a current work practice so that control can be maintained under changing conditions.

Approach#

Compare the current practice with the proposed blueprint. Specify:

  • human and AI roles, responsibilities and decision rights;
  • primary and supporting literacies and chosen collaboration modes;
  • workflow stages, information sources, provenance and trust classifications;
  • evidence passed between stages;
  • tool permissions, validation, approval and escalation points;
  • risks, failure modes, safeguards and checkpoints;
  • how the practice should adapt when conditions change.

Justify the design with Module 1 evidence, new experiments, stakeholder feedback or relevant analogous observations.

Core Project Evidence#

At overview level, the project should contain:

  • a protocol and diagnosis of the current practice;
  • the proposed Human-AI Collaboration Blueprint;
  • explicit predictions and at least one meaningful comparison;
  • inspectable evidence from execution, structured walkthroughs or credible review;
  • responsibility and decision logic;
  • analysis of security and trust boundaries;
  • a record of one adaptation stress test;
  • interpretation, residual risks and a personal theory of control.

Possible supporting artefacts include diagrams, interaction patterns, source strategies, risk registers, logs, outputs, screenshots, traces, retrieval evidence, review notes and stakeholder feedback. The detailed project brief will determine format and scale.

Project Process#

Use the experimental learning loop from Module 1:

  1. Hypothesis: What should improve, and why?
  2. Intervention: What changes in the interaction, data, workflow, responsibilities or system?
  3. Observation: Which outputs, traces, decisions, behaviours or feedback will show the effect?
  4. Interpretation: What does the evidence support, and where are its limits?

Include at least one meaningful baseline-to-redesign comparison. Examples are current versus redesigned practice, weak versus checked grounding, unclear versus explicit responsibilities, or a tool-dependent process versus a transferable control pattern.

Include one adaptation stress test. Change a consequential condition, such as the model, source, policy, permission, human role or risk threshold. Observe which controls still work, which fail and what needs to change.

Personal Theory of Control#

Explain how reliable, responsible collaboration can be achieved and adapted in the chosen context. State:

  • what needs controlling and which observable signals show whether control is maintained;
  • which interventions help and on what evidence;
  • how work is divided, where human judgement is decisive and what evidence permits reliance or action;
  • which assumptions, boundaries and residual risks limit the explanation;
  • what should change when tools, data, policies, people or working conditions change.

For example, a participant working with support replies might develop this provisional rule:

I use judge scores to locate replies for review after checking the judge against human ratings. A high score alone does not authorise a customer response. If the support policy changes, I update the context and repeat those checks before relying on the scores.

This is an illustration, not an additional requirement. A project would need evidence for the rule, its limits and the proposed response to change. Express the theory so that another person can question, test and revise it.

Bridge from Module 1#

Use the portfolio to connect diagnosis, redesign, testing, adaptation and transfer. Module 1 supplies bounded experiments; Module 2 asks how their findings hold together in a sustained redesign of one work practice.

Assessment Orientation#

Assessment focuses on adaptive mastery. A strong project presents:

  • a meaningful problem, clear diagnosis and coherent blueprint;
  • justified collaboration modes and use of primary and supporting literacies;
  • credible evidence, inspectable artefacts and interpretation of outcomes;
  • source trust, permissions, validation, approval and security boundaries;
  • a meaningful adaptation stress test;
  • responsible handling of human, organisational and ethical constraints;
  • residual risks and transfer beyond one product;
  • a personal theory of control grounded in the evidence.

This document provides orientation. The detailed Module 2 brief will specify submission format, scope and assessment criteria.