Case Studies

Case Studies: What happened, and what can we learn?

About this resource · How to use it · Learning engineering connections

What is this resource?

A source-linked collection of documented cases, grouped by domain to compare failures, interventions and capability designs. Each case preserves its evidence qualifications and scope limits.

How should I use it?

  1. Choose a domain and read two cases. Compare the system conditions, actions and outcomes before drawing a lesson.
  2. Follow the sources and disclosures, then propose a new context where the explanation might fail. The linked casebook provides the wider reading edition.

How does this connect to LENS and learning engineering?

Contrasting cases support conditional judgment about human and system performance. LENS asks which capability, information, authority or support changed what people could do; a historical association alone does not establish that a proposed intervention will work elsewhere.

Compare what people could do, what the system made possible, and what mattered to the people affected. Some cases document learning engineering. Others show where it could directly address a failure, or help you distinguish a capability problem from a technical or governance problem.

19 selected cases across 7 domains. Drawn from Capability Matters: A Casebook. Download the revised reading edition below.

Read the memo: Why LENS uses a casebook.

To try a teaching simulation, visit Experiments. To develop and test an intervention, explore Prototypes.

Keep all three in view.

Human
Who gained or lost agency, opportunity or well-being? Whose experience is missing from the record?
Science
What evidence supports the proposed learning explanation, and what other causes could explain the outcome?
System
Which tools, responsibilities, interfaces or conditions enabled the result or prevented recovery?
Human flourishing · Learning sciences · Systems engineering →

How to read the cards: “What failed” and “What worked” describe the reported outcome in the case, not a judgment of the people involved. Paired cases invite comparison; they do not establish that one factor caused an outcome. Competency tags identify the LENS practices exercised, while topic domains identify the setting. Open a case to examine its sources and limits.

1 Healthcare & Patient Safety

What failed

Therac-25

1985 – 1987 Potential learning engineering response

A false underdose reading and a one-key retry left a therapist without a dependable way to recognize the hazard.

Read the case →
Capability in practice

Keystone ICU / Pronovost Checklist

2004 – present Capability design

Michigan ICUs reported lower infection rates after an intervention combining a checklist, stop authority, supplies, education and feedback.

Read the case →
What worked

I-PASS Handoff Bundle

2014 Documented learning engineering

Residents practiced a structured handoff, and the incoming clinician checked what they understood.

Read the case →

2 Education, Training & the Learning Workforce

What failed

inBloom

2014 Failure diagnosis

A shared student-data platform wound down amid disagreement over what families, teachers and districts would gain or risk.

Read the case →
What worked

Cognitive Tutor / Carnegie Learning

1990s – present Documented learning engineering

An algebra tutor used a model of mastery to guide practice. A school-level trial shows what the full curriculum achieved.

Read the case →
Capability in practice

Latin American Learning Analytics Project

2017–2020 Capability design

Universities in Chile and Ecuador used local evidence to shape learning analytics tools and governance.

Read the case →

3 Aviation & Aerospace

What failed

Boeing 737 MAX / MCAS

2018 – 2019 Potential learning engineering response

Two crashes exposed the need to test pilot response against the actual behavior of the MAX under simultaneous failures.

Read the case →
What worked

Crew Resource Management & CAST

1981 – present Documented learning engineering

Crew Resource Management made coordination a skill crews could practice and demonstrate. CAST supplied a separate industry hazard and evidence process.

Read the case →

4 Defense & National Security

What failed

USS Vincennes & Iran Air Flight 655

1988 Potential learning engineering response

A correct radar track did not produce a correct command decision. Follow the information from sensor to crew to command.

Read the case →
What worked

DARPA Digital Tutor

2009 – 2014 Documented learning engineering

A 16-week tutor produced strong knowledge-test results in one small Navy comparison. The practical results varied by task.

Read the case →
What worked

Navy Surface Warfare Readiness Reform

2018–2024 follow-up Documented learning engineering

Career assessments and fleet feedback gave the Navy new ways to examine whether training transfers to work at sea.

Read the case →

5 Industry, Energy & Enterprise Systems

What failed

Knight Capital Trading Loss

2012 Potential learning engineering response

One server missed the update. The signals and controls that could have contained the loss failed.

Read the case →
Capability in practice

Toyota Production System / Andon Cord

1950s – present Capability design

A Toyota worker can signal a problem while the line keeps moving. What happens in the rest of the work cycle determines what the team can learn.

Read the case →
Capability in practice

INL Turbine-Control Upgrade

2014 Capability design

Operators helped refine a digital control interface before formal validation.

Read the case →

6 Disaster Prevention & Recovery

What failed

Deepwater Horizon

2010 Potential learning engineering response

Conflicting readings, no agreed test procedure, and a separate mechanical barrier that failed.

Read the case →
What worked

Navy SUBSAFE

1963 – present Documented learning engineering

After Thresher’s loss, the Navy built recurring training and independent audits into submarine safety certification.

Read the case →

7 Algorithms, Governance & Public Systems

What failed

UK Post Office Horizon Scandal

1999 – 2015 Potential learning engineering response

Horizon could show a shortfall that did not exist, and branch operators could not record a dispute in the system.

Read the case →
What worked

Estonia X-Road

2001 – present Context and comparison

A shared exchange layer solves one coordination problem and raises another: how do agencies move off it safely?

Read the case →

Why LENS uses a casebook

The casebook puts the argument of Capability Matters into practice. What people can accomplish depends on what they know and can do, and on what the system makes possible. The cases ask where a gap came from, what could change it, and what evidence would justify acting. Ask who benefits, who carries the burden, and who has the authority to act or contest a decision.

I-PASS shows how resident practice, faculty feedback, and a structured handoff work together. In the INL turbine-control case, operators used a simulator to expose interface problems before formal validation. Failure cases let readers design and test a response to a documented gap while keeping the proposed response distinct from what happened.

Rwanda RapidSMS shows why sending a referral and completing it require different evidence. The Latin American Learning Analytics project shows how local inquiry shaped advising tools and pilots. Navy Surface Warfare Readiness Reform asks how to tell whether assessed skills persist and transfer to supervised work at sea.

The LENS approach moves from Understand through Map, Design, Build, Instrument, Experiment, and Evaluate, then returns to Understand. Refine throughout. Pick a case, define the work, map what people and systems must provide, and specify a result that would make you revise the design.