- Finite resources
- Surprise is fundamental
- Change never stops
- the behavior of a system is function of interactions of parts/components that make up the system.
- Interactions happen across multiple scales.
- What is a system is a matter of perspective of the analyst who chooses the boundary
- Goals: acute vs chronic
- Plans: efficiency vs thoroughness
- Complex systems: optimality vs resilience
- Adaptive histories
- Precarious present
- Resilient future
What happens in adaptive cycles? (see below)
Systems are more precarious than we think
- different units in our tangled, layered network have models
- those models can be wrong
???
Ability to anticipate bottlenecks ahead & take actions to prepare.
- anticipate how we might get crunched/squeezed
- generate readiness to respond
Work to rule/role
It is necessary/good to pursue optimality.
Handling surprises regularly
Well-adapted activity occurs with a facility that belies the difficulty of the demands resolved and the dilemmas balanced.
Adaptations hide what was adapted.
Competence envelope continually expands
Model surprise is always possible.
Optimizing within the competence envelope always makes you more brittle in the borderlands.
Joint function of:
- graceful extensibility
- optimality
Combine how well we do:
- far from the boundaries (far from risk of saturation)
- near the boundaries (risk of saturation is high)
- Studies of resilience in action
- regularities
- patterns
- descriptive
- explanation
- charting how changes reverberate through multi-role, multi-eschelon networks
- each response to change by some unit triggers adaptive responses across other units
- pressing need intersects with apparently powerful capability
- point of change reverberates through the network
- experience of adaptive shortfall (brittleness)
- responsible people act as generic adaptive and learning stop-gap to fill shortfall
- resulting in kludges (workarounds)
- new opportunities for innovation
- adaptive florescence
stalls <---reverb---> florescence
- change in one unit opens up new possibilities in other units
- the degree that changes in one area tend to recruit or open up beneficial changes in many other aspects of the network, which opens up new opportunities across the network
- Decompensation (complexities in time)
- Desynchronization: Working at cross-purposes (complexities over scales)
- Getting stuck in outdated models (complexities in learning)
- exhaust capacity to adapt
- all units have bounds
- saturation
- "going solid"
- breakdown occurs when challenges grow and cascade faster than responses can be decided on and deployed to effect
- Can I keep up with change?
- How much margin is left
- Readiness to
- Deploy existing
- Mobilize new
- Generate new (the future)
- Teams are no longer coordinated
- fragmentation (silos)
(See later)
- every good controller is a model of system being controlled
- pursuing targets in face of variation
- model surprise
- learning loops
- world changes, but system stuck in what were previously adaptive strategies
- proactive learnings vs. getting stuck
- oversimplification
- fixation
- Proactive learning without waiting for collapse
- two kinds of learning
- change your base envelope/fitness
- graceful extensibility: change how stretch at boundary
- cushion of potential actions & additional resources that allow system to continue functioning despite unexpected demands
- controlling risk of saturation
- critical parameter for control anticipation
- there are regularities of what happens at borderlands
- can be measured
- anticipate vs decompensate
- synchronize across units vs work at cross purposes
- roles, levels, proactive learning before failures vs. stuck in stale behaviors
From The theory of graceful extensibility: basic rules that govern adaptive systems
- Boundaries are universal
- Surprise occurs, continuously
- Risk of saturation is monitored and regulated
- Synchronization across multiple units of adaptive behavior in a network is necessary
- Risk of saturation can be shared
- Pressure changes what is sacrificed when (adapting how you adapt)
- Pressure for optimality undermines graceful extensibility
- All adaptive units are local
- Perspective contrast overcomes bounds (individual perspectives are limited)
- Reflective systems continually risk mis-calibration (you think your model is better than it is)
- uses its knowledge & capabilities to adjust its behavior in changing circumstances in pursuit of goals
- responsible for goals
- experience consequences of success/failure
Human systems as adaptive units are:
- intentional (have goals)
- reflective (have a model of themselves)
- Solution to the tragedy of the commons
- Proposed by Elinor Ostrom
- In between command hierarchy and fully decentralized
- Network of interdependent units operating over different ranges
- strategy for
- empower decentralized initiative
- coordinate over emerging trends to meet priorities
- dynamic interplay as situations evolve & as other units adapt
- multiple centers
- each has partial authority/autonomy
- Example: Balinese water temples (unintended consequence of Western intervention was pest outbreaks)
- Reciprocity: "I will help you when you're crunched"
- need to incentivize working across roles
- Pick a trigger
- Who does the trigger affect?
- Adaptive capacities can be regulated
- No UAB can have sufficient range of adaptive behavior to regulate risk of saturation by itself
- Some UABs can monitor & regulate capacity for maneuver of other UABs in response to changes in the risk of saturation
Treat each adaptive unit response as a trigger.
Look at waves of change.
- acute, immediate, slowly building
- anomalous?
- opportunity?
- bottleneck
- who reacts to it?
- who notices it?
- what does it mean for that role/UAB?
- How did they change?
- recognize risk of saturation
- adjust capacity to maneuver
- how hard is it to stay in control?
- struggle for fitness is ongoing
- have to look at network
- example: heart as adaptive unit
- Optimality-brittleness (Doyle)
- Acute-chronic, faster-better-cheaper pressure (Woods)
- Efficiency-thoroughness (Hollnagel)
- Bounds on perspective (Woods)
Five tradeoffs:
- Optimality vs resilience
- Efficience vs thoroughness
- Concentrated vs distributed
- Local vs distant views
- Acute vs chronic goals
When pressure/load is low, lots of good operating points, easy to find them.
As load goes up, harder to find good operating points
Every system is continuously stretched to operate at capacity.
Lecture 9, 1:21:08
- Tangible experience with surprise
- Unease that the system is more precarious than they think
- Push initiative down
- Reciprocity: sync work across units of adaptive behavior (units help each other)
- Work to align goals across multiple units
- Invest in graceful extensibility
- recognizes decompensation is a risk
- know there are dragons
- talk about the dragons
- talk about how the dragons change
- recognizing what are good at handling dragons
- talk about: how do I deal with cascading problems?
- if I wait until a need is definitive, it is already too late
- practice & simulation when don't see dragons that often
- regardless of how successful they have been in the past
- past success does not mean future will be the same
- system is poised more precariously
- what worked yesterday won't
- ready to adapt how you adapt
- decentralize initiative
- empower local initiative to unit of action in the world
- see this in places where:
- surprise is really, really tangible
- failure to handle surprise is really, really bad
- example: military history
- upper eschelons:
- monitor how well handlings things
- reinforces area of success
- filling in gaps in order to achieve high-level goals
- Ostrum's heuristics
- synchronize work across units of adaptive behavior
- I'm sensitive to how hard you're working to stay in control when you're at risk of saturation
- I'm going to do something to relax or assist you when that happens
- I'm going to extend your ability to perform in those conditions
- I'm going to build graceful extensibility in our relationship
- most organizations when they have signs of trouble:
- put additional pressure on you to work to role
- here's your standard and what you should be doing
- here's the scores you have to hit
- here's the criteria you must meet
- more and more pressure to meet those scores/criteria
- that creates role retreat
- don't risk spending time/energy reaching across role boundary
- don't assist other units of adaptive behavior
- undermines synchronization & coordination
- guarantees an increase in working at cross-purposes
- role specification only includes standard forms of interaction across roles
- recognize there are multiple goals, conflicts, reprioritization required
- help them work out how to align goals
- modify what goals are most important for the current circumstances we're dealing with
- in an accident: get away from productivity goals, focus on critical safety goals as a situation degrades (e.g., Apollo 13)
- "leadership" is a stopgap: using generic adaptive capacity of people to make up for the fact that you don't have a system that works together well
- need to coordinate upper & lower eschelons closer to points of action in the world: polycentric governance
- reserve needs to be an effective force
- Surprise is about preparation
- there are regularities about surprise
- systems can be adapted to those regularities about surprise
- need capabilities in advance
- we can get better at building adaptive capacity
- Connectivity
- Sensors
- Automation/autonomy
- People (human goals & expertise)