Life-like agent-based programming
Power& tradition holding back progress Contents Overcome reactionaries

Emergent agent-based programming

Summary
This presentation applies complex adaptive system (CAS) agents to computer programming. 
Introduction
Complex adaptive systems (CAS(p.17)) that have developed through evolution, including biological cells & our bodies, use emergent(p.24) agent(p.25)-based programming.  Being goal oriented, and able to learn and evolve, agents also provide an alternative paradigm for software development
  • Software development has a problem(p.3) - its applications are inflexible, of limited quality, have no inbuilt capacity to learn about and explore their environments, it provides limited support for emergence and it is wasteful
    • My friend wonders why her Skype service is not running.  Eventually she reboots her PC in frustration.  The responsible Microsoft engineers are unaware of her problems
    • My source editor is task specific and reflects the structure of the whole tool chain
    • My Firefox browser fails to restart a previously aborted session.  Its a typical brittle piece of software failing to cope well with the unexpected.  
  • Manufacturing overcame analogous problems
    • Many of the cars that came off the production line had been damaged during assembly and had to be reworked
    • Production line workers see that the car they are working on has a damaged in assembly but they had no way to stop the line, and no one to tell about the cause of the problem
    • Such problems were overcome by the application of Just in time(p.5) methods, and in particular the effective use of agents
    • Assembly line operators (human agents) were encouraged to use their experience to detect assembly problems, stop the line so that the cause could be identified and fixed and the work plan improved. 
  • My programming framework Smiley(p.7) which uses agent-based programming to avoid the types of problems outlined above is reviewed
  • Opportunities(p.8) and challenges(p.14) to introducing agent technologies in software development
  • Complex adaptive systems CAS(p.17) theory provides a framework for understanding CAS agents

- next page -> (1 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
rhlcontents


<- previous page - - next page -> (2 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
The software development problem:

Software development has a problem - its applications are inflexible, of limited quality, have no inbuilt capacity to learn about and explore their environments, it provides limited support for emergence(p.24) and it is wasteful
  • Programmers still write software programs as instruction sequences
    • Goals of development are only indirectly represented in the implementation 
    • The programming methodology typically uses procedures with parameters  
      • Parameters undermine emergent reuse of software
    • Typing schemes aim to support correct deployment of the procedure hierarchy
      • Typing undermines emergence and exploration of the programs environment
    • Phenomena(p.24) are not represented in the programming methodology
  • Requirements, specification, design and operation activities are discrete processes
    • Frequently the source instructions become the only accurate specification and design reference
  • Any change to the software's requirements will need programmers to make equivalent changes to the source code
    • Software is not designed to respond adaptively to any situation
      • My source editor is task specific and reflects the structure of the whole tool chain
  • The runtime code is brittle
    • My Firefox browser fails to restart a previously aborted session
  • Software engineers must provide adaptations but typically have limited visibility of the users operations
    • My friend wonders why her Skype service is not running.  Eventually she reboots her PC in frustration.  The responsible Microsoft engineers are unaware of her problems


<- previous page - - next page -> (3 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
The problem (continued):
  • Designers are required to successfully transform requirements into operational software
    • Top down functional design transforms requirements into specialized data structures and pre-specified instruction flows
      • Knuth's famous book 'Algorithms + Data structures = programs'
      • Need for a programmer to adjust the data structures to extend the program raised the cost of reuse
      • Event services must be explicitly deployed with care taken to avoid deadlocks
    • Alternatively Object orientation packages the data and algorithms into objects
      • Object encapsulation makes the representation invisible from outside the object
      • Designer must decide how to decompose a system into objects
      • Object design patterns support the designer in selecting appropriate objects
        • Application object decomposition still designed specifically for each situation
        • Designer still decides at design time what scenarios are covered and how they are represented and processed
        • Events framed as messages between objects
      • Limited goal action separation is supported through virtual functions and inheritance
  • There is little transparency of when instruction flows are inappropriate to the situation they are operating in
    • Learning(p.32) from problems and potential improvements is not supported by the runtime code
  • Sequential software coding paradigm is brittle, and limits flexibility
    • Parallel code operations can facilitate combining alternative strategic approaches


<- previous page - - next page -> (4 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Introduction to Just in time (JIT) manufacturing methodology
  • JIT was developed to improve flexibility(p.33), quality, and reduce waste in manufacturing
  • Removed errors, reduced transportation, stock and waiting times
  • Involved and empowered all employees
  • Encouraged iterative design
  • Cross functional participation helped enrich the perceptions applied in design for manufacture
  • Views manufacturing operation as a network of process and operational flows
    • Operations add value by transforming inputs into desired outputs - non JIT manufacturing focused on operations aiming to
      • Reduce labor costs
      • Automate
    • Processes link the operations together and add no value so JIT aims to minimize them - JIT studies process flows and aims to
      • Restructure to maximize effectiveness of linked operations
      • Reduce the costs of the processes


<- previous page - - next page -> (5 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Just in time (JIT) methodology improved quality and reduced waste in JIT manufacturing(p.5)

One radical aspect of JIT was its assumption that trained operators motivated by quality goals would provide powerful, immediate checking, and focused solutions

JIT made operators into agents(p.25):
  • Given goals and asked to solve problems
  • Empowered to perform Shewhart cycles(p.32)
Now we will look at our agent-based programming(p.8) system


<- previous page - - next page -> (6 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Adaptive web framework codelet testing system (Smiley)
Smiley is an emergent(p.24) agent(p.25)-based programming system (methodology(p.12)) with:
  • Cascades of schematically(p.23) controlled application agents using a stack of emergent services
    • Genetic code processing complex
    • Operon based codelet deployment via Copycat codelets
  • Supported by a chemical environment simulation
    • Association infrastructure - coupling Copycat codelet based forces with particular structural molecular groupings
    • Bonding of chemical objects into molecular groups driven by the action of bonding and grouping Copycat codelets
      • Barriers to bonding and grouping due to absence of necessary phenomena within the local chemical elements
    • Implementation of Douglas Hofstadter and Melanie Mitchell's Copycat architecture including
      • Coderack(p.10) - platform for scheduling and executing Copycat codelets
      • Slipnet - network of concept relationships
    • Infrastructure supporting programmatic operations on generalized data structures(p.9)
    • Infrastructure supporting multiple Workspaces
    • Infrastructure binding phenomena to keywords
The chemical infrastructure, and service codelets support the aggregate operation of flexible goal based schematic agents(p.8)
Agents execute indirectly represented by cooperating codelets scheduled bottom-up by the Coderack, or top-down by already 'executing' agents


<- previous page - - next page -> (7 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Programming with flexible goal based agents
Emergent(p.24) agents(p.25) offer the potential to make programming systems sensitive to their situation, flexible and goal oriented.  They are:
  • Flexible by
  • Goal oriented by using schematic(p.23) association between:
    • Modeled(p.27) situation, and executed strategy
      • For example - Model codelet predicts niche is present if resource concentration increases
      • It finds resource does increase so strategy codelet is signalled
    • Signal and aggregate(p.11) of agents deployed
  • Sensitive to situation
    • Environmentally aware through indirect associations of phenomena with structurally integrated agents
    • Decisions being based on schematic state
    • Successful schemata being retained in offspring
    • Operon state controlled deployment of agents
Programming methodology(p.12) associates building blocks with codelets to program agents by aggregation


<- previous page - - next page -> (8 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Generalized data structures
By using the byte stream as a fundamental building block of its services and applications UNIX supported emergence of cooperative applications and utilities integrated by pipes.  Smiley(p.7) similarly integrates its codelets and services around a list based data structure building block which implements the Copycat Workspace(p.7)

all codelet(p.12) based agents(p.25) can interact with the data structures deployed in the Workspace
  • They may not be interested but that is their decision
Smiley's generalized data structure - is a list based 'object':
  • Labeled - every 'object' has a keyword label
  • Operon associated - sequences of labels can be actively associated with programmatic forces
  • Has associated structural properties and forces which describe the physical environment through a 'conceptual' network of keywords
  • Represent instances of program activity (state) as a Workspace(p.7) network of 'actual' labeled objects
  • Represents schematic structures(p.23) as
    • Instance of a partial Workspace(p.7) network
    • Conceptual associations of operons with programmatic forces


<- previous page - - next page -> (9 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Parallel exploration of alternatives
The active programming elements of each agent(p.25) (Copycat codelets(p.12)) compete for resources: 
  • As they achieve their local goals they obtain positive feedback, increasing the priority of their flow of processing, and deploying data structures which again encourage further action
  • To achieve this the infrastructure supports scheduling of 'codelets', which must perform each task as a series
  • The infrastructure based on Douglas Hofstadter's Copycat Coderack architecture schedules each codelet based on its salience (randomly picking from a normalized set of priorities*number of instances of a particular type of codelet)


<- previous page - - next page -> (10 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Signalling cascades 
Successful Copycat codelet(p.12) builders perform actions including issuing signals.  Signals cause the genetic apparatus to deploy signalled schematic structures and hence indirectly their associated codelets.  The signals are:
  • Supported and relayed from initiator to final targets, via cooperating codelets,
    • They initially penetrate the barrier between the operational Workspace(p.7) and the nuclear Workspace routing the signal
    • First to the 'polymerase' complex
      • Mapped to signalled operon group schematic strings
    • The mapped outputs are transported back across the nuclear membrane and on to deployment in the Workspace
  • Indirectly bound via schematic(p.23) association
    • Mapped by a genetic code processing 'polymerase' complex into a target operon
    • Mapped signals sent to target operon's associated aggregations of codelets which together perform the signalled goal
A cascade of signalled codelet deployments typically results from each signal due to operon grouping


<- previous page - - next page -> (11 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Agent programming methodology
Four programmed aspects define each building block of an aggregate agent(p.25):
  1. Declaration of the conceptual schematic string(p.23) and its associations in the Copycat Slipnet(p.7)
  2. Declaration of schematic string representing signal(p.11)-to-operon association in a Copycat Workspace(p.7)
  3. Declaration of schematic operon sub string which when signalled will be deployed into an active Copycat Workspace
  4. Codelet Perl source code associated by Slipnet declaration (1) with the active deployed schematic string (3)
    • Each codelet is composed of separately scheduled stages:
      • top-down scout - optional, as Hofstadter architecture(p.7)
      • scout - optional, searches for items of interest to it within the proximate environment
      • evaluator - required to check environment to see if builder can run
      • builder - performs the agent's component action including signalling the nucleus
      • inhibitor - optional but if present will inhibit evaluator until conditions change to suppress it
      • inspector - optional but if present will support inspection of the codelet's progress
    • Each codelet stage can submit the next stage to the Coderack
    • Builder codelets can signal(p.11) the nucleus
State changes of the agent are restricted to those that can be represented by changes to the deployed schemata
Aggregate agents emerge(p.24) from:
  • Temporal chain of programmed signals linking operational codelets
  • Cascading of deployed codelets generated by operon grouping


<- previous page - - next page -> (12 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Agent models
Models(p.27) are implemented in Smiley(p.7) as schematic(p.23) agent(p.25) aggregates where:
  • Builder action is to verify a hypothesis attribute
  • Multiple model codelets can be deployed in parallel to evaluate alternative scenarios
  • Smiley allows Coderack concept activation to influence the salience of modeled codelet signals(p.11)


<- previous page - - next page -> (13 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Challenges of agent-based programming
Emergent(p.24) agent(p.25)-based programming is not without its challenges:
  • Indirection, and parallelism are
    • Computationally costly compared to direct deployment of serial code with jumps
    • Difficult to operationally comprehend
  • Associations are
    • Costly
    • Adaptive web framework (AWF) associations are resolved at run time in competition to the operation of the codelets
  • Agents decide the strategies used to achieve the goals
But these challenges spotlight both the
  • Reasons why real world agents use the architectures they do
  • Awesome capabilities of cellular infrastructure amplifiers(p.34)


<- previous page - - next page -> (14 of 15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Emergent(p.24) agent(p.25)-based software architectures have the potential to model(p.13) their behavior and learn from problems
  • Use of goals telling agents what must be done rather than how to do it enables flexible problem solving
  • Inclusion of this key aspect of JIT should improve the quality and flexibility of software based systems


<- previous page - (15)
 

Power& tradition holding back progress Contents Overcome reactionaries
Backup slides follow

- next page -> (backup 16 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Complex adaptive systems
  • Complex system:
    • Constrained by emergence(p.24)
    • Obeys a small set of rules
    • Represents environmental state
    • State variables interdependent based on the nature of the links between the state variables
    • Hurricane: component air and water molecules transiently present; It obeys physical laws; energy within hurricane can interact with (reflect) external environment
  • Complex adaptive system (CAS):
    • Uses captured energy to operate on plans and strategies
    • Executed by emergent(p.24) schematic agents(p.25)
  • CAS appear chaotic unless viewed with a representative model
  • Agents adapt to other local agents



<- previous page - - next page -> (backup 17 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Signals and sensors
  • CAS(p.17) interact with their environment via sensors detecting signals, and emergent(p.24) agents(p.25) (1) performing actionsstructurally enhanced state agent array
  • Signals are phenomena(p.24) generated by the action of agents
  • A sensor's state will change when it interacts with an environmental phenomena such as a signal
  • As long as a sensor has differential state depending on what phenomena it is experiencing, evolutionary selection(p.30) can drive the emergence of advantageous associations
    • Evolution can capture beneficial strategies within the schematic plan(p.23)
    • An agent can associate each alternative state with a separate strategy
  • Signals can be multi-modal and require very rapid processing
    • Front-end processing sensors, such as eyes (2) and ears (3), which include structurally enhanced state(p.19), and massive parallelism emerges(p.24)
    • Networked associations can then represent the environmental situation (4), responses (5) and their valuations



<- previous page - - next page -> (backup 18 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Structurally enhanced state
  • Emergent(p.24) agents(p.25) can only respond to signals(p.18) if they are able to represent them as state changes
  • When the signals are complex (view of animal walking in a wood), and multi-modal (sound and light), or occur very rapidly (falling object) it may be necessary to prophylactically setup a structure to represent the alternative states
  • The occurrence of a complex modal signal can then be represented by a change to the structure
  • The mechanism is very flexible
    • Structural linkages between agents can represent multi-modal signals
    • Pre-deployment of agents can support detection of rapidly changing signals
  • The structural state representation must be schematically(p.23) defined and modeled(p.27) by schematic agents


<- previous page - - next page -> (backup 19 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Flows: control and partitioning
  • Objects accelerated by a force, channeled by phenomena(p.24) in their path create persistent emergent(p.24) patterns
  • CAS(p.17) can transform stored energy into forces that
    • Channel resource objects into flows
    • Partition flows
  • Network(p.21) of CAS agents(p.25) gather, channel flows of resources 
    • Resources transformed by agents into intermediate/end products
  • Energy consuming transportation agents can be integrated with agents monitoring(p.18) and controlling the flow
  • Buildup of intermediate products detected/used to inhibit early stage transport and operations
  • CAS agents provide an adaptive, emergent mechanism to control and partition flows


<- previous page - - next page -> (backup 20 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Network effects
Network effects sustain improved efficiency of flows and control(p.20) between connected agents(p.25)
  • Examples of network effects include:
    • Water molecules hydrogen bonding to one another to form liquid water. 
    • Routers linked together enabling the operation of the Internet
    • Web pages linked together by HTTP and the Internet enabling operation of the World-Wide-Web.  
  • Network effects include:
    • Positive returns of hub connections
      • Increasing benefits for each element connected to a hub.  
      • Sexual selection(p.31) amplification of flows of genes and memes
    • Transition from chaos to order - as the number of elements connected increases the likelihood of most elements being connected increases asymmetrically, enabling flows and control(p.20)
    • Well-connected network is robust to node failures
    • Infrastructure amplification(p.34) improves flows, and enables control


<- previous page - - next page -> (backup 21 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Network effects (continued)
  • Network effects can emerge as agents(p.25) connect to one another
  • Topology introduces constraints and other properties to the network 
    • One additional link can significantly effect the system
  • Closely connected agents can cluster 
    • Outside the cluster weak links/highly connected hub agents connect other parts of the network
  • In agent networks the resources transported through the network include
  • System dependent on continued resource flows


<- previous page - - next page -> (backup 22 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Introduction to schematic structures
Schematic structures are constructed from simple building blocks, such as nucleic acids or the letters of the alphabet.  These blocks are able to link together into groups, lists and networks, and directly or indirectly associate with agents(p.25)

  • Deployed based on a fundamental building blocks (1) which can: schematic string
    • Provide brick for agents, including genetic operators(p.28), to build structures
    • Represent a data value and be
    • Replicated and
    • Linked into lists (2)
    • Integrate descriptors, and addressable labels
      • Descriptors can be interpreted broadly by agents as for example valuations or epi-genetic control signals(p.18)
    • Associatively network
  • Any schematic sequence of basic elements can be interpreted by an agent as a signal (3) to perform a specific act
    • Agents position on the schematic structure represents a state and so the local sequence can be a signal
    • Becomes an association (4) between signalled event and the agent's action
    • Actions can include modeling(p.27) of a situation


<- previous page - - next page -> (backup 23 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Emergent layers - phenomena -> physical -> chemical -> adaptive epiphenomena
  • Low level phenomena underlie the world: strings, quarks etc.
  • Physical and chemical effects result from the interactions of these low level physical phenomena
  • Simple predictive laws emerge from the characteristics of the physical phenomena observed within a bounded environment
    • Atoms, molecules, electrons obey Bohr's predictive laws as long as the phase is bounded by everyday scales of measurement
  • Bohr atoms, molecules, electrons are epiphenomena
  • Epiphenomena emerge as conceptual theories that predict the outcomes of interactions of real phenomena operating on a platform such as our world
    • Have re-combinable building blocks
    • Aggregate 
    • They are [in]directly active
  • Adaptive [epi]phenomena, also emerge from the interactions of low level phenomena, but they include


<- previous page - - next page -> (backup 24 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Agents: schematic aggregates generating action
Agents are emergent(p.24) aggregations of catalysts.  The aggregations are formed schematically(p.23) associating programmed control of catalytic actions with different goals and events
  • Agents - association between goals and actions.  They
  • Perform actions (1) by focusing fundamental physical forces, marshaled as accessible stored energylogical agent
  • Respond to
    • Signals(p.18) from their environment (2)
    • Other agents
    • Their own state (3)
    • From this aggregation goals emerge
  • Constructed from active structures formed from re-combinable building blocks 
  • Utilize a schematic plan(p.23) (4) to indirectly associate the signals with actions
  • Invoked under operon control
  • Form emergently(p.24)


<- previous page - - next page -> (backup 25 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Emergent(p.24) agents process environmental signals(p.18) and their instance of the schematic plan to select actions which they perform
Agents can interpret the schematic structures as they please, but acted on by evolutionary selection(p.30) beneficial associations of signals and actions transform into strategic attributes of the system
  • A labeled element, operon, with associated actions corresponds to a goal
  • Specific goals can be associated with multiple, alternative actions
  • Individual agents can be deployed with a full complement of operons, but different, allelic, sets of associated actions
  • Filtering of labels and other tags provides a flexible control of flows(p.20)
    • Agents can filter, and integrate the flows of schematic elements to perform genetic operations(p.28)
  • Genetic operations provide the basis for learning, by correlating success in replication of schematic structures and the retention of valuable sequences with positive agent actions
    • Requires that agents actions are based on schematic signals
  • Total set of different germ-line schematic structures shared in genetic operations corresponds to the evolving state
  • Ability to replicate schematic structures allows for somatic copies to be augmented with highly specific state descriptors, while germ-line copies are left pristine
  • Multi-cellular agents are able to deploy different states in individual somatic cells


<- previous page - - next page -> (backup 26 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Models provide agents with a schematic view of the world
  • Emergent(p.24) agents(p.25) must be able to differentiate good strategies from bad.  Models assist them with making such decisions
  • They are hypothesis that use success or failure as an indicator of legitimacy
  • Correlation of model and schematic(p.23) agent action allows evolutionary selection to 'rate' models in particular niche situations
  • Agents can extend the evolved rating mechanism by applying a Shewhart cycle(p.32)


<- previous page - - next page -> (backup 27 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Modifications to the germ-line structures are performed by genetic operators 
  • Emergent(p.24) agents(p.25) can include genetic operators which allow reproduction to formally transform the current structures for the next generation
  • Genetic operators can equivalently be deployed via a discrete process called a genetic algorithm
  • Operations include: mutation, crossover, inversion, translocation, deletion, dominance modification


<- previous page - - next page -> (backup 28 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Phenotypic alignment:
Emergent(p.24) agents(p.25) competing in adjacent environmental niches gain fitness from synergistic network effects(p.21) 
Selection pressure re-enforces the process of alignment  
  • Mother's genes can generate agents which support their developing offspring.
    • Support will apply the mother's phenotypic strategies to the development of the offspring, which may be different to the phenotypic action of the offspring's agents 
    • Mother's genotype has aligned with the offspring's phenotype.  More generally,
  • Male genes can express phenotypically signals deployed with their sperm to inhibit a female from accepting further sperm (from competing males)
    • Induces an arms race(p.34) between males & females
  • Multiple genes may contribute to a phenotypic effect 
    • Alleles may compete for effect
    • Genes can be in separate bodies (males & females competing in the arms race above) 
  • The extended effect may be present whenever one organism appears to be manipulating another
Dawkins central theorem "An animal's behavior tends to maximize the survival of the genes 'for' that behavior, whether or not those genes happen to be in the body of the particular animal performing it.  "


<- previous page - - next page -> (backup 29 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Natural selection: collecting situation specific beneficial recipes within an adaptive agent
Evolution: interaction of selection of agents(p.25), variation of schemata, heredity

Structure of schematic plan(p.23):
Schematic specification/control of operational cascades:
  • Schematically represented Darwinian pre-adaptations can recur generation after generation
  • Pre-adaptations become niche specific selective advantages
    • Association between niche and beneficial phenotypic effect is formed
    • Schematic amplifier can emerge


<- previous page - - next page -> (backup 30 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Sexual selection:
 CAS(p.17) agents(p.25) reproduce using genetic operators(p.28) to
  • Integrate the schematic plans(p.23) of each parent
  • Indirectly select genetic structures
  • Particular phenotypic effects
    • Generation of a trait in male offspring and
    • Selection of that trait in female offspring
  • Responsible genes collocated in genomes of both expressive males and females (polygenes)


<- previous page - - next page -> (backup 31 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Shewhart cycles: Walter Shewhart advocated the use of iterative cycles of: Planning, Doing, Checking and Acting to fix issues (PDCA)
  PDCA cycle emerges(p.24) naturally in complex adaptive systems
  • Examples include the cell cycle, cell death programme, cell growth programme, Just-In-Time manufacturing
  • Plans in an adaptive system are schematic structures(p.23)
    • Associated with current environmental state indirectly through modeling(p.27)
    • Models generate predictions of results of performing each strategy
  • Performing, or 'do'ing, a highly valued strategy will result in some change of state of the agent(p.25) and its environment
  • Agent can compare (check) the result against the models prediction
    • Human agents evolved strategies often avoid checking so a required process supports the formation of the habit
  • If reality differs from the prediction the valuation of the strategy can be adjusted


<- previous page - - next page -> (backup 32 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Strategies for success
Schematic aggregation(p.25) results in the emergence(p.24) of goals and strategies - clusters of actions which are beneficial when a particular situation recurs
  • Flexibility - requires that different generic situations can be recognized and responded to appropriately
    • Schematic retention of good strategies allows these to be deployed when the situation repeats
  • Personality aligns around focus on the group and concreteness - Individualist Artisans and group oriented Guardians are interested in concrete activities
    • Individualist Rationals and group oriented Idealists are more conceptual
  • Guardian morality - punishment by the whole group, enforced by the Guardians, on those who cheat on their commitments to the group
  • Centralization - when agents migrate to the most connected regions of the network their influence increases
  • Prophylaxis - key nodes in a network can be supported by the close attention of multiple cooperating agents
    • Before they come under pressure
    • Often deterring competitors from moving to those nodes


<- previous page - - next page -> (backup 33 of 34)
 

Power& tradition holding back progress Contents Overcome reactionaries
Strategies for success (continued)
  • Infrastructure amplification - uses structures that reduce startup costs of operations to catalyze flows.  Enzymes, banks and roads are examples
  • Evolved amplification - deploying rules and inducing a collection of agents to act
    • So that the inducer gain significantly
    • Even as the induced agents benefit a little
  • Platform - focusing agents on a network hub with
  • Structurally enhanced state - Compound signals, can be represented, and processed by cooperating networks of agents
    • General data structures typically limit what a single agent may represent
    • Representation is defined by the schematically controlled deployment of agents in the network under evolved pressure
  • Bundling - Shared schemata allow strategies to be deployed cooperatively, and differentially, so that specialized agents can operate as a super organism
  • Disruption - with limited energy and constraining models hub network agents can be induced to prioritize strategies
    • Guarantee their nodes collapse responding to newly joined low cost nodes


<- previous page - (backup 34)