Tropos Research Overview: RWTH Aachen - PowerPoint PPT Presentation

paul2
slide1 l.
Skip this Video
Loading SlideShow in 5 Seconds..
Tropos Research Overview: RWTH Aachen PowerPoint Presentation
Download Presentation
Tropos Research Overview: RWTH Aachen

play fullscreen
1 / 38
Download Presentation
Tropos Research Overview: RWTH Aachen
424 Views
Download Presentation

Tropos Research Overview: RWTH Aachen

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

  1. Tropos Research Overview: RWTH Aachen M. Jarke and G. Lakemeyer RWTH Aachen

  2. History of Aachen’s Approach DAIDA : Metadata linking IS development phases, development of Telos, first ConceptBase prototype (1988) COMPULOG : logical foundations of meta modeling, meta formula handling by partial evaluation (Jeusfeld 1992) NATURE/CREWS : application to perspective resolution in requirements engineering and business modeling (Nissen 1997, Peters 1997), extension to business cooperation (Kethers 2001) and warehoues/e-commerce (Quix 2002) TROPOS : dynamic modeling of social networks in coooperation with sociologists and entrepreneurs

  3. common meta model entity relationship notation data flow notation viewpoint 1 viewpoint 2 common reality Multi-Perspective Cooperative Modeling : Metamodels as Goals For small processes [Nissen et al.; IEEE SW 96]: --> metaplan + repository formalization For large processes [Rose, CACM 12/98]: --> layout generation + synchronization analysis in

  4. Regional Networks in the US Emilio J. Castilla, Hokyu Hwang, Ellen Granovetter, and Mark Granovetter (2000): Network of Silicon Valley Entrepreneurs

  5. Network Theory Approaches • Cluster Theory (M. Porter) • Innovation and productivity by Coopetition • Social Capital (Bourdieu, Putnam) • Reputation mechanisms (individual / organisation) • central role of trust • our TCD approach (Trust-Confidence-Distrust) • networks not a hybrid • distrust is not negation of trust • successful combination high trust - high distrust !

  6. Entrepreneurship Network Route 128,dependencies according to 40 interviews Strategic investor Local Cluster Information Technology Investment Banker Angel Group Angel Limited Partners Venture Capitalist Entrepreneur lawyer customer Supply/ Service Consultant/ Mediator Team E-Center 50k- Competition E-Lab Licensing Office Faculty Alumni

  7. Modeling Organisation Networks with Tropos: A Trust-Centered Approach M. Jarke and G. Lakemeyer Aachen University of Technology other contributors: L. Ellrich, C. Funken, M. Meister, Freiburg G. Gans, S. Kethers, R.Klamma, T. Vits, Aachen V. Wulf, Fraunhofer FIT

  8. Goal- and Scenario-Based Continuous Requirements Management scenario generation for validation/ refinement observation focus/ goal discovery goal/requirement refinement/negotiation new model initial model change specification animate reverse analysis future scenario change envisionment current scenario change implementation capture legacy integration existing system new system ... extended to Social Networks

  9. Social Networks • an autonomous form of coordination that combines advantages of two traditional mechanisms • markets: flexibility, adaptive to competition • hierarchies: long duration cooperation • trusting cooperation of autonomous, but interdependent agents (here: organizations) • small-business B2B commerce, virtual org‘s • entrepreneurship networks

  10. Overview 1. Role of Trust/Distrust in Agent Networks 2. A Multi-Perspective Modeling Approach 3. The Tropos(AC) Prototype 4. Conclusions

  11. expectation of risky investment reciprocity individual trust : likelihood of reciprocity violation Coleman: p G > (1-p) L (similar: DAI community) game theory: iterated prisoner‘s dilemma eBusiness: reputation mechanisms network confidence : strategic vulnerabilities, balanced inter-dependency networks Trust

  12. An Example: Seminar Organization Eric Yu‘s i*,here: the strategic rational model

  13. Distrust is not just negation or absence of trust Luhmann 1988: functional equivalence + distrust spiral Kern 1997: inject controlled amount of distrust Lewicki et al. 1998: high-trust / high-distrust networks can be highly successful Implications for RE disparity between motive and action (latency) positive function in networks (watchful trust) tendency for distrust aggregation must be contained, e.g. by network rules and traceability (institutionalized distrust) A Basis for IT-supported continuous RE ! Distrust

  14. The TCD Model : a delicate balance for network success

  15. Overview 1. Role of Trust/Distrust in Agent Networks 2. A Multi-Perspective Modeling Approach 3. The Tropos(AC) Prototype 4. Conclusions

  16. Enhance Viewpoint Resolution Methods: But which Viewpoints are needed? Expectations + Reality ... communicated in negotiations among agents Reaching goals requires certain actions Speech Acts Plans depend on (and influence) Trust, Confidence, Distrust

  17. Multi-Perspective Agent Model: The TCD Kernel

  18. Multi-Perspective Agent Model: Structural Perspectives

  19. Multi-Perspective Agent Model: Action Perspectives

  20. Speech Act Perspektive Action Workflow [Medina-Mora et al. 1992, Schäl 96]

  21. Delegation within Speech Acts

  22. Speech Act Modeling: TCD influences delegation structure No trust: Pro‘s: no risk Con‘s: narrow competence

  23. Speech Act Modeling: TCD influences delegation structure Trust: Pro‘s: broader competence Con‘s: delayed reaction

  24. Speech Act Modeling: TCD influences delegation structure Confidence: Pro‘s: improved competence and efficiency Con‘s: higher risk

  25. Speech Act Modeling: TCD influences delegation structure Confidence plus Distrust: Pro‘s: focused risk management Con‘s: monitoring effort

  26. Extending i* for a Trust-Based Approach • Task preconditions and postcondition • synchronization of interdependent agent-plans • Trust, confidence, and distrust as subj. probabilities • as part of preconditions they control task selection • A simple (linear) model of time • monitoring expectations (distrust, network rules)  Facilitates the mapping into an executable plan language

  27. Organizer´s Plan with Confidence ClientOrganizer Speaker

  28. Watchful Confidence

  29. Overview 1. Role of Trust/Distrust in Agent Networks 2. A Multi-Perspective Modeling Approach 3. The Tropos(AC) Prototype 4. Conclusions

  30. From Extended i* to Congolog programs • Congolog: A logic-based plan language • offers control constructs like sequences, if-then-else, while, recursion, and concurrency • allows reasoning about actions with a solution to the frame problem (Reiter 1991) • comes equipped with an interpreter forsimulation Example: proc sem_org <c,l,t.request(c,l,t)  /* client c requests seminar l to be held at time t */ if nettrust(c)  NWThreshold then (commit(c,l,t); organize(c,l,t)) else /* call a proc. to find speaker first and then commit */ endproc

  31. OME ext. i* Golog prog. .pl file static model Telos file Eclipse Golog Int. Java Interf. TCP/IP CBase Simulator (Java) History The System Architecture

  32. A Restaurant Domain everything_ok rate_ good good_rating rate_ meal eating_ finished bad_rating rate_ bad something_ not_ok serve_ meal meal meal eat go_to_ restaurant ready cook start work meal_ chosen accept_ order order_ received choose_ meal trust_high_ enough guest chef waiter exogenous

  33. The Extended i* Editor OME guest waiter chef exogenous

  34. Congolog Code generated from ext. i* /*Agent procedures */ proc(agent_guest, conc( interrupt(eating_finished=true , decomp_rate_meal), interrupt(ready=true , decomp_go_to_restaurant))). proc(agent_chef,interrupt(order_received=true , cook)). proc(agent_waiter, interrupt(meal_chosen=true , decomp_work)). /*Decomposition of complex Tasks*/ proc(decomp_work ,[ conc(accept_order , serve_meal), work]). proc(decomp_go_to_restaurant ,[ conc(choose_meal , eat), go_to_restaurant]). proc(decomp_ rate_meal ,[ conc([rate_bad , rate_meal] , [rate_good , rate_meal])]).

  35. Simulator 1

  36. Simulator 2

  37. Simulator 3

  38. Conclusions • worked with sociologists to look into the dynamics of how dependencies and strategic rationals change due to the evolution of trust and distrust in a social network • Introduced methodology based on i*, ConGolog, and speech acts to model and simulate trust-based inter-actions in agent networks. Havefirst prototype. • To evaluate our model, we are studying networks of start-up companies, in Germany and the US (MIT). Also look at b2b applications (M. Schoop, T. List). • Interested in understanding relationships to other formalizations (model-checking) and applications. Formal joint project?