Towards a Simulation Tool for Evaluating Dynamic Reorganization

Towards a Simulation Tool for Evaluating Dynamic Reorganization of Agent Societies V. Dignum, F. Dignum, Utrecht University L. Sonenberg, University of Melbourne V. Furtado, A. Mello, University of Fortaleza

Outline § § § § Main issues Motivation Organizations and change Reorganization typology Reorganization requirements The VILLA simulation environment Conclusions

Main issues in Reorganization 1. When 2. Why 3. What • • Behavior Structure 4. Who 5. How • • External to system: re-design Within system: dynamic – Decision, authority

Motivation § Need for organization – Stability – Cope with uncertainty § Need for reorganization – Answer to change • environment, objectives, population – Flexibility

Need for reorganization § Need for organization: achieve stability § Reorganization means loss of stability § Need for reorganization – Answer to environment changes – Answer to population changes – Answer to objective changes § Reorganization decision depends on – organizational utility and success. – individual utility and success

Organizational Utility § Organizational Success: ability to bring assets to bear, recognize and take advantage of opportunities. – Ex: MAS perspective • • Interaction success: how often result in desired aim? Role success: how often agents achieve role goals? Structure success: how well are global goals achieved? Combined utilities of agents in organization § Utility: determined based on – Current success – Expected success – Cost of reorganization

Individual Utility § Success and utility are different for each agent – Dependent on own goals and resources – Dependent on social attitude § Role utility to agent success determines participation in organization

Moments of reorganization – when and why? § From organizational theory § Timing (timeliness) – Proactive: prepare for expected change – Reactive: adjust after change § Intention (resiliency) – Offensive: gain advantage – Defensive: ensure survival [Evans, 1991]

Reorganization manouvres reactive proactive Timing Intention offensive defensive pre-emptive protective -Expected future -Competition through innovation exploitive -Taken after event -Capitalize opportunities -Expected future -Limit damages corrective -Taken after event -Prevent more damage -Ensure continuity

Focus of Reorganization – what? § Behavior change – Temporary, ‘local’ to one population – A new agent joins the MAS – An agent leaves the MAS – Interaction pattern instantiation § Structural change – Permanent, valid across populations – Organizational Self Design – Structural Adaptation

Means for reorganization – how? § Decision – concerns how reorganization decisions are reached. – Relates to the decision-making style. § Authority – Concerns what aspects can be changed by who in the organization – Relates to the C 2 Model

Reorganization Decision – who? § Externally imposed – Agents have no influence on reorganization – occurs through system redesign § Role based – Command-driven: the agent does not make any decisions on how to pursue its goals, and some other agent has authority over it – Locally autonomous/master: The agent makes decisions alone and may or not have control over other agents § True consensus – Agent works as a team member, sharing decision making control equally with other agents [Barber, Martin, 2001]

Reorganization Authority – what? § C 2 Model – Command: authority and responsibility to determine the objectives of the organization. Can update the social structure of the organization. – Control: authority to specify and modify detailed plans for achieving objectives. Authority to modify interactions and behavior. § C 3 Model: C 2 plus – Communications: collection and sharing information about the environment, the state of the organization, the state of the achievement of objectives, and the state of execution of the plans. • Meta-communication [Galey, 1987, Tidhar, Sonenberg, 2001]

Dynamic reorganization styles Behavior Structure collaborative Shared control Shared command directive Decision Authority Role-based control Role-based command

Requirements for reorganization § Observation – Identify patterns – Evaluate current response possibilities – Generate options, predict outcomes, understand effects § Organizational requirements § Agent capabilities

Organizational requirements § Reorganization success – Timeliness – Adaptance/consensus – Resiliency § Depend on organizational form/domain – Hierarchy: role-based reorganization decision? – Network: shared reorganization decision? – Market: favors behavior change?

Agent capabilities Authority collaborative Structure Shared Control Shared Command • no memory • reasoning • meta-communication directive Decision Behavior Role-based control Role-based Command • no memory • memory: only role • reasoning: only role • no meta-communic.

Simulation Aims - 1 § Full theory of reorganization is more than what can be studied in one simulation § Agent behavior depends on – Own state and environment state – But also on the organizational structure – Organizational structure is thus not just a component of the environment § Organizational elements considered: – Type of goal (simple to complex) – Roles (many agents, one agent) – Interactions (communication protocols, dole dependencies)

Simulation Aims - 2 1. Identify match of organization structure to environment characteristics 2. Adaptation to (drastic) changes – Structural vs. behavioral – Role-directed vs. collaborative 3. Communication requirements to reason about change – Also, reasoning with limited knowledge

The VILLA environment § Aim: community survival § Creatures – Gatherers: can collect (limited) food individually – Hunters: can hunt (large amounts of) food in groups – Others: consume food, can grow to become Gatherers or Hunters – Head: observe and change society

VILLA: Activities § Simulation takes a number of runs (days) § In each run: – Eat • • If food available Collectors eat more than others If not eat, health decreases If health = 0, then creature dies – Collect • Gatherers: individual function on health • Hunters: groups’ function on health and size – Move • Hunters must move to form group

VILLA setup

VILLA without reorganization

Evaluation of VILLA § Influences on health: – Role typology – Role capabilities § Results from evaluation of non reorganization situation: – Food stack decreases a lot at beginning – Need to introduce delay in adaptation – Others average health seems to be good indicator for reorganization – Need to evaluate time interval, not time point

Evaluation of VILLA (no reorg) G 17 6 H 0 11 0 0 0 9 8 0 17 0 9 8 8 8 7 7 7 9 9 9 6 5 4 3 8 5 6 6 5 7 3 5 6 7 10 11 11 11 0 4 3 2 4 3 7 3 2 1 1 1 2 3 O Comments Gatherers survive with 100% of health. All creatures die because amount of food is not sufficient to keep a good health level. All creatures die. Only in cases when the hunters get together very early some creatures survive. Hunters keep others alive if food stack is very high (more than 10000) All creatures survive more than 100 TICs. However, food stack must be 900 to allow Hunters to get together within 500 ticks. Very good society but depends on the probability of Hunters to get together. Stable society with health 80%. However some Others will die. Good and stable society with health greater than 80% All creatures die Very good society Good society Very good society with health 95% but instable if Hunters are isolated. Very good society with health in 95% but instable if Hunters are isolated. Good society but very instable if Hunters are isolated.

Evaluation of VILLA (no reorg) G H O Prob. Gather Prob. Hunter Comments 9 5 3 9 10 Instable Society, depending on hunters’ aggregation. 9 6 2 9 10 Instable Society, depending on hunters’ aggregation. 9 5 3 15 18 Stable society independent of hunters aggregation. Reach 100% and food stack increase. 9 6 2 15 18 Stable society independent of hunters aggregation. Reach 100% and food stack increase. 9 0 8 15 18 Stable society independent of hunters aggregation. Reach 100% and food stack increase. 5 0 12 15 18 Minimum number of gatherers for supporting other life. 0 17 0 15 18 With the increasing of prob. Hunters always still alive and keep society good 8 5 4 15 18 Health society before was 80% now 100%. 7 7 3 15 18 Stable society independent of hunters aggregation. Reach 100% and food stack increase. 7 6 4 15 18 100% “ 7 5 5 15 18 100% ” 7 5 5 18 20 100%

Reorganizing Societies § Behavioral change: – If food stack < 250, increase gather power by 1 – Reorganization delay is 100

Reorganizing Societies § Structural change: – If food stack < 250, create 1 gatherer (from Others) – Reorganization delay is 100

Conclusions § First step towards a model of reorganization § Identification of characteristics of reorganization § Requirements for reorganization – Different organizational types – Challenges for agent capabilities

Future work § Empirical study – How systems react to different reorganization forms – Further evaluation/development of VILLA § Reorganization methodology – Conditions and requirements – Determine choices § Formal model for reorganization? – Organizational utility – Reorganization cost