Coordination of Complex Socio-technical Systems: Challenges and Opportunities
1 like214 views
The document discusses the coordination of socio-technical systems (STS), highlighting the challenges and opportunities in bridging the socio-technical gap. Key themes include the need for self-organization, addressing abstraction gaps, and leveraging argumentation-based coordination to improve accountability and transparency. It also presents approaches such as speaking objects and a model for organizing knowledge, emphasizing the integration of socio-cognitive aspects in engineering practices.
![In practice?
❖ STS examples
❖ Internet of Things deployments
❖ Computer Supported Collaborative Work (i.e. WfMS)
❖ Social Networks
❖ Gap examples
❖ Amazon Alexa funny accidents
❖ Electronic Medical Records failures [Park et. al. 2012]](https://image.slidesharecdn.com/talk-180626094851/85/Coordination-of-Complex-Socio-technical-Systems-Challenges-and-Opportunities-3-320.jpg)
![❖ BIC bottom line:
❖ practical behaviour is a means for communicating
❖ no specialised signal needed (i.e. speech acts)
❖ Tacit messages: implicit communicative meaning conveyed by actions
❖ “turn on lights while leaving home” –> “somebody is in” (to potential intruders)
❖ “process X does action A” –> “actions based on A now enabled” (synchronisation)
❖ taxonomy with examples in [Castelfranchi et. al. 2010]
Observation: BIC](https://image.slidesharecdn.com/talk-180626094851/85/Coordination-of-Complex-Socio-technical-Systems-Challenges-and-Opportunities-13-320.jpg)
![Self-organisation: biochemical coordination
❖ (Bio)chemical coordination bottom line:
❖ chemical-like reactions as coordination rules
❖ interplay of reactions running locally originates global patterns
❖ May implement many coordination “patterns” (like OO design patterns)
❖ basic: aggregation, spreading, repulsion, …
❖ composite: digital phermones, gossiping, foraging, …
❖ catalogue with methodology in
[Fernandez-Marquez et. al. 2013]
simulated. Consequently, self-organising design patterns
are best exploited during the design phase of a chosen
methodology.
The design patterns come into play during the design
phase, which we propose to split into three distinct steps
(Fig. 3): (1) the choice of design patterns is made during an
early phase of design. Self-organising design patterns serve
to identify the problem to solve as well as to determine the
appropriate solution to bring to the problem. In particular,
they help determining the boundaries of each problem and
its corresponding solution provided by the pattern; (2)
during a refined phase, actual entities and their dynamics
are defined. The patterns’ dynamics serve to refine the
model and to identify the entities and their precise inter-
actions, individual responsibilities and to anticipate the
emergent behavior; (3) finally, during the simulation step,
successfully to different self-* systems. By analysing their
behaviours, we identified common lower-level mechanisms,
some of them basic (atomic) and other composed of basic
ones. As a result, we classified the patterns into three layers.
The basic mechanisms that can be used individually or in
composition to form more complex patterns are at the bottom
layer. At the middle layer, there are the mechanisms formed
by combinations of the bottom layer mechanisms. The top
layer contains higher-level patterns that show different ways
to exploit the basic and composed mechanisms.
Figure 4 shows the different design patterns collected in
the catalogue and their relations. The arrows indicate how the
patterns are composed. A dashed arrow indicates that it is
optional (e.g. the Gradient Pattern can use evaporation, but
the evaporation is not necessary to implement gradients).
This classification aims at listing existing mechanisms
from the literature, identifying their own boundaries (i.e.
when one mechanism stops, and when another starts), their
inter-relations and the recurrent problem they solve. For
example, Gossip has been applied to many works in dif-
ferent ways, but all implementations share the fact that
Analysis
Design
Implementation
Verification
Test
Requirements
Early Design
Phase
Refined
Design Phase
Simulation
Design Pattern Choice
Transition rules
Environment
Computational model
Methodology
Design Phase Design Patterns
Fig. 3 Design patterns within the design phase of SO methodologies
HighLevel
Patterns
Composed
Patterns
BasicPatterns
ForagingFlocking
GossipDigital Pheromone
MorphogenesisQuorum Sensing
Evaporation AggregationRepulsion
Gradients
Chemotaxis
Spreading
Fig. 4 Patterns and their relationships](https://image.slidesharecdn.com/talk-180626094851/85/Coordination-of-Complex-Socio-technical-Systems-Challenges-and-Opportunities-17-320.jpg)
![Argumentation: coordination
❖ Argumentation-based negotiation bottom line:
❖ argumentation framework as coordination rules
❖ arguments as complex info driving negotiation (i.e. strategy behind bid)
❖ Not only negotiation!
❖ many different dialogue games with own goals, requirements, engagement rules, …
❖ agents engage in dialogues depending on goal (i.e. joint planning, info collection, …)
❖ reference categorisation in [Walton, Krabbe 1995]](https://image.slidesharecdn.com/talk-180626094851/85/Coordination-of-Complex-Socio-technical-Systems-Challenges-and-Opportunities-21-320.jpg)
![References
❖ [Park et. al. 2012] Park, S. Y., Lee S. Y., Chen, Y.: “The effects of EMR deployment on doctors’
work practices: A qualitative study in the emergency department of a teaching hospital”
International Journal of Medical Informatics (2012)
❖ [Castelfranchi et. al. 2010] Castelfranchi, C., Pezzullo, G., Tummolini, L.: “Behavioral implicit
communication (BIC): Communicating with smart environments via our practical behavior and
its traces” International Journal of Ambient Computing and Intelligence (2010)
❖ [Fernandez-Marquez et. al. 2013] Fernandez-Marquez, J.L., Di Marzo Serugendo, G., Montagna,
S., Viroli, M., Arcos, J.L.: “Description and composition of bio-inspired design patterns: a
complete overview” Natural Computing (2013)
❖ [Walton, Krabbe 1995] Walton, D., Krabbe, E. “Commitment in Dialogue: Basic concept of
interpersonal reasoning” Albany NY: State University of New York Press (1995)](https://image.slidesharecdn.com/talk-180626094851/85/Coordination-of-Complex-Socio-technical-Systems-Challenges-and-Opportunities-40-320.jpg)
Coordination of Complex Socio-technical Systems: Challenges and Opportunities
- 1. Coordination of Socio-technical Systems Stefano Mariani Department of Sciences and Methods of Engineering Università degli Studi di Modena e Reggio Emilia Reggio Emilia, Italy Challenges and Opportunites
- 2. Goal ❖ Fact: IT systems and society are NOT isolated systems ❖ Socio-technical Systems (STS) as the result of their interaction ❖ Issue: socio-technical gap when STS peculiarities overlooked ❖ Aim: fresh look on STS engineering, coordination perspective ❖ NOT exhaustive, NOT optimal: experience on directions worth exploring :) ❖ focus on “core”, foundational mechanisms
- 3. In practice? ❖ STS examples ❖ Internet of Things deployments ❖ Computer Supported Collaborative Work (i.e. WfMS) ❖ Social Networks ❖ Gap examples ❖ Amazon Alexa funny accidents ❖ Electronic Medical Records failures [Park et. al. 2012]
- 4. ? Where are people in IoT?
- 5. Where are people in WfMS?
- 7. Challenges: self-organisation ❖ STS have emergent properties ❖ can be designed? how? ❖ how to asses them? simulation? run-time? ❖ Awareness is key (“who is doing what”) ❖ what about scale? privacy? ❖ IT platform should adapt ❖ should users know why? ❖ should users know expectations?
- 8. Challenges: abstraction gap ❖ Abstraction gap 1: goals vs. actions ❖ humans reason in term of goals (“I want to chill”) ❖ devices understand actions (“switch music on”, “dim lights”, “light fireplace”, …) ❖ Abstraction gap 2: situations vs. measurements ❖ human reason in terms of situations (“is this place on fire?”) ❖ devices understand measurements (“is temp > X?”, “is smoke detector triggered?”, …) ❖ How to reconcile? ❖ more intelligent devices? (or more stupid people? )
- 9. Challenges: accountability ❖ The fear of algocracy ❖ (“filter bubble” effect, employment chance, insurance profile, healthcare access, …) ❖ not an issue on its own ❖ Lack of accountability is! ❖ “who to blame”? “what’s going on”? ❖ tradeoff: transparency vs. privacy
- 10. Opportunities: observation ❖ Observation-based coordination ❖ well known example: stigmergy ❖ less known: Behavioural Implicit Communication (BIC) ❖ Foundational elements: ❖ environment as mediator of (inter)action ❖ visibility of actions and their traces (~ effects on environment) ❖ notion of locality (for observation)
- 11. Observation: example ❖ Main outcome: self-organisation (by emergence) ❖ agent X does action A0 causing modification M0 ❖ agent Y sees M0 and does A1 causing M1 ❖ agent Z sees M1 and does A2 causing M2 ❖ … ❖ If Ai = “sort brood” –> Mi = “pheromone smell” => brood sorting :) ❖ local = “move item from here to there if similar items there” ❖ global = partial clustering of items based on similarity
- 12. Observation: evolution ❖ Further steps: ❖ cognitive stigmergy = stigmergy + symbolic reasoning ❖ BIC = cognitive stigmergy + actions + awareness ❖ Symbolic reasoning: traces have meaning ❖ Actions: made observable likewise traces ❖ Awareness: agents know they are observed by others
- 13. ❖ BIC bottom line: ❖ practical behaviour is a means for communicating ❖ no specialised signal needed (i.e. speech acts) ❖ Tacit messages: implicit communicative meaning conveyed by actions ❖ “turn on lights while leaving home” –> “somebody is in” (to potential intruders) ❖ “process X does action A” –> “actions based on A now enabled” (synchronisation) ❖ taxonomy with examples in [Castelfranchi et. al. 2010] Observation: BIC
- 15. Opportunities: self-organisation ❖ Self-organising coordination ❖ decentralised approach to coordination (no coordinator) ❖ well known examples: birds flocking, ants foraging, wolves sorrounding prey, … ❖ less known (?): (bio)chemical coordination ❖ Foundational elements: ❖ actions sensitive to context (situatedness) ❖ notion of locality (for interactions) ❖ (often) probabilistic decision-making (or stochastic = probability changes with time)
- 16. Self-organisation: example ❖ Main outcome: adaptation (by emergence) ❖ if local context is C0 then do action A0 with p00 = 1 ❖ if local context is C0 then do action A1 with p01 = 0.8 ❖ if local context is C1 then do action A1 with p11 = 0.2 ❖ … ❖ If C = “info (un)known” + A = “store / forward” => gossiping :) ❖ local = “probabilistically forward or not info based on context” ❖ global = broadcast resilient to failures and network (re-)configuration
- 17. Self-organisation: biochemical coordination ❖ (Bio)chemical coordination bottom line: ❖ chemical-like reactions as coordination rules ❖ interplay of reactions running locally originates global patterns ❖ May implement many coordination “patterns” (like OO design patterns) ❖ basic: aggregation, spreading, repulsion, … ❖ composite: digital phermones, gossiping, foraging, … ❖ catalogue with methodology in [Fernandez-Marquez et. al. 2013] simulated. Consequently, self-organising design patterns are best exploited during the design phase of a chosen methodology. The design patterns come into play during the design phase, which we propose to split into three distinct steps (Fig. 3): (1) the choice of design patterns is made during an early phase of design. Self-organising design patterns serve to identify the problem to solve as well as to determine the appropriate solution to bring to the problem. In particular, they help determining the boundaries of each problem and its corresponding solution provided by the pattern; (2) during a refined phase, actual entities and their dynamics are defined. The patterns’ dynamics serve to refine the model and to identify the entities and their precise inter- actions, individual responsibilities and to anticipate the emergent behavior; (3) finally, during the simulation step, successfully to different self-* systems. By analysing their behaviours, we identified common lower-level mechanisms, some of them basic (atomic) and other composed of basic ones. As a result, we classified the patterns into three layers. The basic mechanisms that can be used individually or in composition to form more complex patterns are at the bottom layer. At the middle layer, there are the mechanisms formed by combinations of the bottom layer mechanisms. The top layer contains higher-level patterns that show different ways to exploit the basic and composed mechanisms. Figure 4 shows the different design patterns collected in the catalogue and their relations. The arrows indicate how the patterns are composed. A dashed arrow indicates that it is optional (e.g. the Gradient Pattern can use evaporation, but the evaporation is not necessary to implement gradients). This classification aims at listing existing mechanisms from the literature, identifying their own boundaries (i.e. when one mechanism stops, and when another starts), their inter-relations and the recurrent problem they solve. For example, Gossip has been applied to many works in dif- ferent ways, but all implementations share the fact that Analysis Design Implementation Verification Test Requirements Early Design Phase Refined Design Phase Simulation Design Pattern Choice Transition rules Environment Computational model Methodology Design Phase Design Patterns Fig. 3 Design patterns within the design phase of SO methodologies HighLevel Patterns Composed Patterns BasicPatterns ForagingFlocking GossipDigital Pheromone MorphogenesisQuorum Sensing Evaporation AggregationRepulsion Gradients Chemotaxis Spreading Fig. 4 Patterns and their relationships
- 19. Opportunities: argumentation ❖ Argumentation-based coordination ❖ well known example: agreement technologies ❖ less known (?): argumentation-based negotiation ❖ Foundational elements: ❖ argumentation framework (reasoning over arguments) ❖ rational agents (i.e. stay on topic) ❖ arbiter (i.e. decide winning argument)
- 20. Argumentation: example ❖ Main outcome: accountability ❖ agent X makes assertion A (“S is the state of the world”, “I want resource R”, …) ❖ agent Y challenges A (“State is S’ for sensor Z”, “Resource R is already mine”, …) ❖ agent X defends itself (“Z is faulty”, “Agent W is lying”, …) ❖ … ❖ To win debate, agents have to disclose information ❖ transparency = argumentation / negotiation rules are known ❖ accountability = faults and malicious behaviours spotted and ascribed
- 21. Argumentation: coordination ❖ Argumentation-based negotiation bottom line: ❖ argumentation framework as coordination rules ❖ arguments as complex info driving negotiation (i.e. strategy behind bid) ❖ Not only negotiation! ❖ many different dialogue games with own goals, requirements, engagement rules, … ❖ agents engage in dialogues depending on goal (i.e. joint planning, info collection, …) ❖ reference categorisation in [Walton, Krabbe 1995]
- 22. Outline: 3rd opportunity Abstraction gap Algocracy Arg. based Coord. (negotiation+)
- 23. Approaches: Molecules of Knowledge ❖ Main idea: ❖ exploit users’ interactions to continuously and spontaneously (self-)organise information ❖ promote aggregation of related information and diffusion to interested prosumers ❖ Pillars: ❖ biochemical coordination –> computational model ❖ behavioural implicit communication (BIC) –> interaction model Mariani, S. (2016) “Coordination of Complex Sociotechnical Systems: Self- organisation of Knowledge in MoK” Artificial Intelligence: Foundations, Theory, and Algorithms
- 24. MoK in one slide ❖ MoK system overview ❖ network of compartments where seeds continuously and spontaneously inject atoms ❖ atoms aggregate into molecules, diffuse to other compartments, gain/lose relevance, and so on ❖ these processes are enacted by reactions executing within compartments and influenced by enzymes and traces ❖ enzymes and traces are left within compartments by catalysts while performing their activities information repository sources of information atomic information composite information coordination lawsreification of actions actions’ side effects sw agents or human users
- 25. MoK: peculiarities ❖ Reactions leverage decentralisation and situatedness to promote self- organisation ❖ contextual to information local to their compartment and can only affect neighbours ❖ scheduled according to dynamic rate expressions inspired by natural chemical reactions ❖ few “foundational” reactions detected through simulation ❖ Enzymes and traces exploit the BIC theory ❖ make agents aware of what others are doing ❖ environment pro-actively acts to improve coordination of agents’ activites
- 26. MoK: Information Management ❖ Citizen journalism scenario ❖ MoK-coordinated platform for retrieving, assembling, sharing news stories ❖ while users carry out their activities, MoK processes self-organise information ❖ In particular: ❖ (user action) whenever users mark relevant info… ❖ …MoK attracts similar one from neighbours (system re-action) Mariani, S. and Omicini, A. (2015) “Anticipatory Coordination in Socio-technical Knowledge-intensive Environments: Behavioural Implicit Communication in MoK” Advances in Artificial Intelligence, Lecture Notes in Computer Science
- 27. MoK: Information Management ❖ Squares are compartments ❖ Coloured dots are info ❖ Coloured flags/arrows are enzymes/traces ❖ From time to time clusters or simlarly coloured info appear ❖ Everything based on users’ interactions!
- 28. Outline: 1st approach Awareness Emergence Adaptation BIC + Biochemical Coordination MoK STS engineering
- 29. Approaches: Speaking Objects ❖ Main idea: ❖ sensor and actuator devices will be able to assert complex situations about the state of the world and to autonomously pursue goals ascribed to users or designed for the system ❖ perceptions –> assertions & actions –> goals ❖ Pillars: Lippi, M., Mamei, M., Mariani, S. And Zambonelli, F. (2017) “Coordinating Distributed Speaking Objects” International Conference on Distributed Computing Systems Knowledge Representation & Ontologies Commonsense Machine Learning Goal Oriented Computing Human Computer Interaction Argumentation-based Coordination expresses high-level goals, desiresprovides high-level explanations, goals high-level situations plan of activities and optimization semantic situations semantic commands complement information, provides constraints commands to actuators raw sensor data translates goals into actions, discuss and coordinate activities. Create coherent stories of what happened user interaction user input
- 30. Coordination protocols Speaking Objects in one slide ❖ Speaking Objects overview: ❖ speaking objects jointly collect information about the state of the world and assert them to whom it may concern ❖ hearing objects collectively plan what to do in response to the ever-changing situations perceived by speaking objects ❖ conversational coordination happens via argumentation between speaking and hearing objects ❖ information seeking, inquiry, discovery, persuasion, negotiation, deliberation dialogues are re-interpreted under the coordination perspective sensor devices Actuator devices
- 31. Speaking Objects: peculiarities ❖ Decentralised coordination by leveraging opportunities for negotiation ❖ Embraces “humans-in-the-loop” by enabling interaction in natural language ❖ Deals with trust and algocracy by making explanations and justifications of decision making available and amenable of inspection and interpretation ❖ Dialogue types and conversation moves as foundational mechanisms
- 32. Speaking Objects: Traffic Control ❖ Intersection management scenario ❖ vehicles equipped with an array of speaking and hearing objects, as the intersection itself (i.e., cameras, traffic lights, …) ❖ approaching the intersection vehicles start argumenting with the traffic light about who has the right of way ❖ In particular: ❖ negotiation phase where vehicles try to persuade the traffic light to decide in their favour ❖ dispute settled when the argumentation process finds a solution for which no vehicle has to stop Lippi, M., Mamei, M., Mariani, S. And Zambonelli, F. (2017) “An Argumentation-based Perspective over the Social IoT” Journal of Internet of Things
- 33. Speaking Objects: Traffic Control ❖ Inquiry dialogue for asking right of way ❖ Information seeking for checking ❖ Negotiation + persuasion to converge ❖ Deliberation to give right of way and stop ❖ Shared argumentation rules! A B Ci Cj T Hi T, how long will the light remain green? Hello A, it would last 30 seconds. Hello T , yes I am. Could you keep the green on 30 seconds more? I’m a bit late to work. Is any vehicle reaching the crossroad within a minute? Yes, one vehicle approaching from south in 40 seconds. Hi B, are you crossing straight? Do you mind waiting for one minute? Sorry, I need to cross now for reaching home soon. I see. Could you B turn right and reach home anyway? It’s just to help another vehicle. Sure, that will take about the same time. Thanks T!
- 34. Outline: 2nd approach Emergence Abstraction gap Algocracy Argumentation-based Coordination Speaking Objects STS engineering
- 35. Conclusion: the bottom line ❖ Take aways ❖ engineering STS is hard, harder if socio-technical gap disregarded ❖ technical vs. socio-cognitive perspectives must be taken into account ❖ So, no good news? ❖ we have ways to reconcile the above perspectives ❖ MoK and Speaking Objects are examples stemming from personal experience
- 36. Conclusion: perspective Integration as key as scientists and engineers, we need to find a way to include socio-cognitive aspects in our technical solutions since the very beginning of the design phase, not as an orthogonal dimension to be added later on, or dealt with in an ad-hoc way
- 37. Integration: example ❖ MoK integrates chemical-inspired coordination (technical) with BIC (socio-cognitive) ❖ Speaking Objects integrate goal-orientation (technical) with argumentation-based coordination (socio-cognitive) ❖ They can even work together: ❖ Smart City as a large-scale STS ❖ MoK as the information handling layer ❖ speaking and hearing objects scattered ❖ information evolves according to MoK vision ❖ speaking and hearing objects exploit it to argue
- 38. Conclusion: issues ❖ Despite efforts, there will always be issues ❖ privacy and security clash with awareness ❖ self-organisation clashes with predictability ❖ decentralisation hinders accountability ❖ … ❖ Fine-tuning integration on application needs is of paramount importance
- 39. Questions? Thanks for your attention :) Coordination of Socio-technical Systems Challenges and Opportunites Stefano Mariani Department of Sciences and Methods of Engineering Università degli Studi di Modena e Reggio Emilia Reggio Emilia, Italy
- 40. References ❖ [Park et. al. 2012] Park, S. Y., Lee S. Y., Chen, Y.: “The effects of EMR deployment on doctors’ work practices: A qualitative study in the emergency department of a teaching hospital” International Journal of Medical Informatics (2012) ❖ [Castelfranchi et. al. 2010] Castelfranchi, C., Pezzullo, G., Tummolini, L.: “Behavioral implicit communication (BIC): Communicating with smart environments via our practical behavior and its traces” International Journal of Ambient Computing and Intelligence (2010) ❖ [Fernandez-Marquez et. al. 2013] Fernandez-Marquez, J.L., Di Marzo Serugendo, G., Montagna, S., Viroli, M., Arcos, J.L.: “Description and composition of bio-inspired design patterns: a complete overview” Natural Computing (2013) ❖ [Walton, Krabbe 1995] Walton, D., Krabbe, E. “Commitment in Dialogue: Basic concept of interpersonal reasoning” Albany NY: State University of New York Press (1995)
- 41. Coordination of Socio-technical Systems Stefano Mariani Department of Sciences and Methods of Engineering Università degli Studi di Modena e Reggio Emilia Reggio Emilia, Italy Challenges and Opportunites