Holistic Boundary Analysis
Expand the boundary before optimizing a visible local benefit
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 97%
Harris contrasts a narrow boundary analysis with a holistic one. A narrow analysis asks whether an intervention solves the immediate problem—for example, whether an autonomous weapon keeps one soldier out of danger or automation removes unwanted work. A holistic analysis widens the frame to include responses by rivals, escalation loops, labor transitions, power concentration, and effects on people outside the transaction. The method begins with the tangible benefit, then maps who and what the original frame excluded. It follows reactions through the wider system and compares the full consequence set before acting. The aim is not to reject useful technology automatically, but to make restraint, transition plans, and countermeasures part of the design rather than afterthoughts.
Origin
Extracted from The Diary of a CEO
Core principles
- 01A local improvement can trigger wider system harm
- 02Visible short-term benefits can conceal delayed costs
- 03Competitive responses are part of the decision boundary
- 04Wisdom requires restraint as well as capability
How to run it
- 1
Name the local win
Describe the immediate, measurable benefit the proposed action is expected to deliver.
Pro tip Use the strongest version of the benefit rather than dismissing it.
- 2
Expose the boundary
Write down which people, time horizon, and system interactions the current analysis includes and excludes.
Watch out An unstated boundary can make a partial analysis look complete.
- 3
Map wider effects
Add downstream stakeholders, competitive reactions, feedback loops, displacement, and delayed consequences.
Pro tip Include the response the intervention invites from adversaries and competitors.
- 4
Test escalation paths
Follow plausible chains in which one response triggers another and local control is lost.
Pro tip Mark uncertainty rather than presenting a scenario as established fact.
Watch out A speculative chain should inform safeguards, not masquerade as a prediction.
- 5
Redesign with restraint
Preserve the useful local outcome while adding limits, transition plans, or coordination that reduce system-wide harm.
Watch out Do not scale until the most consequential excluded effects have an owner.
In the wild
Bartlett presents the immediate case that a machine can reduce soldiers' exposure. Harris widens the frame to potential automated responses, escalation between powers, and loss of human control. The broader analysis does not establish that escalation will occur; it identifies consequences the local safety claim leaves out.
→ The decision includes escalation controls instead of measuring only immediate casualties avoided.
A retailer models the productivity gain from robots, then widens the analysis to displaced workers, local demand, retraining time, competitor reactions, and concentration risk. It phases deployment alongside a funded transition plan instead of treating labor effects as external.
→ The company retains efficiency gains while addressing consequences outside the original cost calculation.
Common mistakes
Counting only visible benefits
Immediate measurable gains can dominate attention while delayed or distributed harms remain outside the frame.
Using worst cases as certainties
Widening the boundary should reveal risks and uncertainty, not convert every possible chain into a factual forecast.
Is it for you?
Best for
High-impact technology, policy, automation, or security decisions with interacting downstream effects.
Not ideal for
Simple reversible tasks whose consequences remain reliably contained within a known boundary.
From the transcript
“that's a narrow boundary analysis without looking at the holistic effects”
“The definition of wisdom is having a more holistic picture.”
From the episode
AI Expert: We Have 2 Years Before Everything Changes! We Need To Start Protesting! - Tristan Harris