Architecture blueprint
Map the full sociotechnical system: user behaviour → workflow → data capture → model output → human action → feedback data → governance controls. Mark every reinforcement and balancing loop.
Pathway 3 · Lesson 0190 min
Systems Thinking for AI
Designing AI systems that behave well at scale
Learning Objective
By the end of this lesson, you will be able to: Draw a causal loop diagram for an AI system, identify reinforcing and balancing feedback loops, and apply the Iceberg Model to an AI transformation challenge.
70%+
of AI failures stem from system-level people and process issues
Engage
"Why do AI systems that work perfectly in testing fail catastrophically in production? The answer is almost always systems — not technology."
Explore
Donella Meadows' introduction to systems thinking — the mental models that reveal why complex systems behave the way they do.
Explain
Feedback Loops: The Engine of AI System Behaviour
AI systems don't exist in isolation — they create feedback loops with the humans and organisations around them.
Reinforcing loops (amplifying): Success breeds success. When AI recommendations are trusted and acted upon, users engage more, generating more data, improving the model, increasing trust. This virtuous cycle accelerates AI adoption.
But reinforcing loops also amplify failure: if an AI system produces biased outputs, those outputs influence decisions, which generate biased data, which trains the next model to be more biased.
Balancing loops (stabilising): Governance, human oversight, and error correction act as balancing loops — preventing runaway reinforcing dynamics. Well-designed AI systems have strong balancing loops built in.
📖
The Iceberg Model
Events (what we see) sit above the waterline. Below are Patterns (trends over time), Structures (the systems and policies that create patterns), and Mental Models (the beliefs and assumptions that create structures). Most AI interventions address events. Lasting change requires addressing structures and mental models.
Explore
How to draw and read causal loop diagrams — the core tool of systems thinking applied to organisational challenges.
Elaborate · Worked Example
Systems Map: AI Adoption in an Organisation
Context
Why do some AI rollouts succeed while others stall, even with identical technology?
The Situation
Reinforcing loop 1 (success): AI adoption → time saved → positive experience → more adoption Reinforcing loop 2 (failure): Poor AI outputs → distrust → avoidance → no feedback data → AI doesn't improve → more poor outputs Balancing loop: Champions network → peer learning → skill building → better prompts → better outputs → breaks the failure loop
The Outcome
The systems map reveals that the Champions Network is the highest-leverage intervention — it's the balancing loop that breaks the failure reinforcing loop. This is why culture investment outperforms technology investment.
Try It · Reflect & Apply
Draw a causal loop diagram: Choose an AI system or rollout in your organisation. Identify 2 reinforcing loops (one positive, one negative) and 1 balancing loop. What is the highest-leverage intervention point?
Your notes are saved in your browser for this session.
Evaluate · Knowledge Check
1 / 3What is a reinforcing feedback loop in an AI system?
Builder Engineering Studio
AI System Architecture Studio
Engineering trade-off
Optimising for task speed can reduce critical human review. Decide where deliberate friction is needed to preserve judgment, safety, and learning.
Quality gate
Before deployment, run a failure-mode review: What happens if the model is wrong, unavailable, biased, or confidently uncertain? Who notices first, and what stops harm from cascading?
Evidence to measure
Track adoption, override rate, correction rate, escalation rate, and outcome quality—not just model accuracy or usage volume.
Build task
Create a one-page system map for one AI-enabled workflow and annotate the three highest-leverage intervention points.
External video references
Keep Learning on YouTube
Open in a new tab
Systems Thinking IntroductionEmbedded lesson videoDonella Meadows' introduction to systems thinking — the mental models that reveal why complex systems behave the way they do.Watch · Watch on YouTube Causal Loop Diagrams ExplainedEmbedded lesson videoHow to draw and read causal loop diagrams — the core tool of systems thinking applied to organisational challenges.Watch · Watch on YouTube Introduction to Systems ThinkingPeter Senge × IBMUse this leadership-oriented primer to reinforce feedback loops, learning systems, and systemic intervention points.Watch · 17 min
Human-centred work design
Human Oversight Check
Use this original practice tool to test whether an AI-enabled workflow supports real human judgment, rather than merely placing a person at the end of an automated process.
Self-check
0/5
Start by making the work and decision rights explicit
Calibrate autonomy
Match AI freedom to consequence. Low-consequence drafting may run quickly; high-consequence decisions need explicit human approval.
Keep context visible
Make the objective, inputs, constraints, source material, and current system state easy to inspect before anyone approves an output.
Design active review
Review should require informed judgment, a plausible override, and the authority to act—not a symbolic sign-off after the fact.
Route exceptions clearly
Specify what uncertainty, conflict, missing evidence, or risk looks like—and who receives the escalation with enough context to decide.
Maintain human capability
Protect deliberate practice in the work that people may one day need to verify, challenge, or perform without AI assistance.
Apply it to one workflow
Ready to move on?
Mark this lesson complete and continue to "Embeddings & Semantic Search".