Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Managing Collaboration Effectively Nick Bleech, Jericho Forum Board of Management -with help from Will Harwood, University of Kent - and Wikipedia! Jericho Forum Annual Conference 22 April, 2008 A Jericho Forum ‘Work in Progress’ Introduction • Collaboration: inter-working between people, between systems – Effective collaboration: advancing mutual objectives of the collaborators – Managing collaboration effectively: fostering and maintaining the conditions for effective collaboration (e.g. trust and security) • Collaboration-Oriented Architecture (COA)’s ‘repositories’: – COntrActs: capabilities - relationships - obligations – REPutations: business events - outcomes - performance/ satisfaction of the COntrAct • Collaboration viewed ‘memetically’ Security without trust A B (request,claim,evidence) e.g. • A wants B to run a program P • B only wants to run programs that it believes are safe • request - run this program P • claim C - it complies with your ‘safety policy’ • evidence - proof of P obeys C (testing = partial proof) What is trust? • A ternary relation: A trusts B for action C – Trust is in the same category of concepts as knowledge and belief – To say I trust you is to assert a belief or knowledge about your actions. – Trust means that we believe a system maintains a property. • Trust involves Risk in that you are handing over control of your interests to another – it is used in place of evidence for behaviour. “Dispositional Trust” - willingness generally to trust “Interpersonal Trust” - based on perceived qualities of the person/ thing being trusted Transactional Trust Positive Incentives Social Trust Negative Incentives Decision to Trust Rational-Legal Traditional Charismatic Authority/Control Dependency “System/Environment Trust” - in things/processes within which a trust relationship exists “Dispositional Distrust” - willingness generally to distrust What is trust? What is trust? • The context for trust decisions – Who to trust - identity – Why to trust - entitlements, rights, permissions – Experience/reputation, beliefs, and verifiability • Security problems – What to disclose in order to achieve a desired trust decision (need to tell/ need to know)? – What not to disclose e.g. to preserve privacy/anonymity? – How to communicate and share knowledge in order to reach the trust decision? – How to capture and communicate experience to maintain trust? What is trust? • If your interests encapsulate my interests then I will trust you. • Encapsulation: The realisation of your interests necessarily leads to the realisation of my interests. • To trust you, I need to believe that both: – Your goals encapsulate my goals, and – You are capable of realising your goals (may invoke interpersonal and/or system/environment trust) • Trust is (should be) used when providing evidence is either not possible/feasible or very costly. • Trust is (should be) rational. (Federated) Security with Trust A B request result Bilateral evidence/proofs of behaviour replaced by “identity proofs”, and “assertions” (claims) but trust in principals’ (agents’) behaviour still needed • A wants B to do C • claims – I am A, I am B, A is permitted C at B, … • evidence • credentials for A, B • delegation certificate for C is permitted for A at B ‘Traditional’ Trusted Third Party (TTP) A B Visa/MasterCard/… • Works well in financial setting • TTP is a Risk Absorber - really deferred trust Problem • Tension: – e-business network effect and power of ‘mass collaboration’ (unorganized collaboration) – Versus: the need to manage collaboration effectively – Mass collaboration models look attractive, but don’t seem to advance all parties’ objectives all of the time, e.g. trust and security • Existing TTP constructs are problematic: – Pooled liability, architectural inflexibility – “We don’t use a TTP in the ‘real world’, so why here, why now?” • What practices, structures and incentives need to be resolved? Why should I care? • Real-world problem: many joint ventures, risk sharing partnerships etc. prove difficult to manage – Concepts, models, guidance needed • Collaborations can (should) be of arbitrary span and depth, so what hope for ‘e-collaborations’? • Mass-collaboration gaining popularity in the e-world: – Social networking, wikis etc. – Social networks complement rather than replace more traditional forms of interaction and social mechanisms (see Clay Shirky: Here Comes Everybody, 2008). – So if we accept that trust and security are inherently multifaceted, social networks can’t provide all the trust and security we may ultimately need. • COA can help Genetic Viewpoint • Richard Dawkins’ The Selfish Gene popularized (and advanced) the genecentric view of evolution: ‘bodies are the gene’s way of making more genes’ • Fundamental concepts here are replicators and vehicles (survival machines) – Replicators include nucleic acids notably DNA, which composes genes (base-pair sequences) – Vehicles include people’s bodies, dogs and fruit flies Genes in Action • Kin selection Genes in Action • Kin selection Genes in Action • Kin selection Genes in Action • If a gene “knows” that another body contains a copy of itself then it gets equal benefit from helping the other body reproduce (inclusive fitness) Genes in Action • If a gene “knows” that another body contains a copy of itself then it gets equal benefit from helping the other body reproduce (inclusive fitness) Altruism among Selfish Genes • Dawkins established that mutual trust among gene-copies can evolve thus advancing the goal of inclusive fitness • Thus genes can pursue non-selfish survival strategies that still advance selfish (to the gene) goals • The way to see this is by considering iterated prisoners’ dilemmas Prisoners’ Dilemma B Co-operate Co-operate Defect R,R S,T T,S P,P • Temptation • Reward • Punishment • Sucker A Defect T>R>P>S Nash Equilibrium, worst mutual outcome, but most logical in absence of trust Iterated Prisoners’ Dilemma • Robert Axelrod demonstrated that when various strategies compete in repeated games of the PD, the ‘tit for tat’ strategy produces the best overall outcome: – A: if B cooperated last time, cooperate this time; otherwise defect – Hence parties that can’t otherwise communicate can do so through their actions, and past actions create a ‘shadow of the future’: basis for trust – Dawkins postulates that many genes preprogram this strategy to maximise survival • This also shows that interaction intensity tends to generate more trusting behaviour. • When thinking about trust and ‘trusting behaviour’, iterated PDs help to uncover rational incentives. Memetic Viewpoint • Dawkins extended the replicator/vehicle paradigm as a way to characterize evolutionary models of cultural information transfer – In this viewpoint, memes stand for ideas, concepts, patterns of thought etc. located in the memory • Memes may alternatively be thought of as observable cultural artifacts and behaviours – Some memeticists argue whether ideas are objectively observable within the memory – In semiotics, signs need to be communicated (copied) and interpreted: memes gloss over the interpretation bit! Memes in Action • Memes, like genes, are copied with variation and selection. Only some variants survive, so memes (and hence human cultures) evolve. – Unlike genetic (DNA) replication, meme replication has a high chance of inducing mutations. – Memes replicate by imitation, teaching and other methods, and compete for space in our memories and for a chance to be copied again. • Large groups of memes that are copied and passed on together are called co-adapted (mutually reinforcing) meme complexes, or memeplexes. E.g. religious ideas. Memes Schmemes • Is Memetics pseudo-science? – Advocates point to promising predictive capabilities • E.g. Jon Whitty’s Memetic model of Project Management (PM): essentially self-serving, evolving and designing organizations for its own purpose. • So PM is a memeplex comprising the stories, rules and norms of project practice and experience. – Organizations believe projects evolve a sense of purpose through their mission statements and explicit goals, but these are often organization’s political compromises – Similarly, projects are seen as superior problem-solving tools, but PM lore focuses mostly on why projects fail not why they succeed, which omits consideration of how else ‘success’ could be achieved. Collaboration as Memeplex? • Jon Whitty’s analysis of PM generates insights, so what about collaboration? • The ‘memetic’ viewpoint seeks to identify a collaboration as a memeplex, and the elements that COA defines/implies as memes • In COA, we associate collaborations with – COntrActs (a meme type) – REputations (another meme type) • Insight: view all three as ‘first class citizens’ • Insight: the architecture should foster ‘inclusive fitness’ of its memes. Towards Effective Collaboration • A standard component of corporate strategy is organizational design (OD) – As corporate strategy has evolved to embrace broader goals, social outcomes, and stakeholder values, OD has evolved too. • Contemporary trends include: – – – – Reinventing hierarchies Project-oriented OD Networks (small-world networks, a.k.a. clusters) Guilds (Eli Lilly example) • All these approaches seek to maximize effective collaboration New OD strengths/weaknesses Type Reinventing hierarchies Strength Weakness Expertise prized, e.g. Little control research teams, jazz bands Remarks A.k.a. ‘adhocracies’ Projectoriented Flexible to focus on Inappropriate for clear goals over finite sustaining durations activities Cf. Jon Whitty critique Networks Flexible, durable, multi-organizational ‘Small world’ networks improve on ‘pure’ networks Guilds Combines features of Long term networks and cohesion? adhocracies Can be swamped by interactions, trust may be shallow Eli Lilly model COA’s first class citizens • Definition: in business terms a ‘repository’ is simply a persistent and dependable record of facts – COntrAct repository models ‘static’ bases for collaboration – REputation repository models ‘dynamic’ collaboration execution performance • Implementation expected to be via ‘repository as a service’ (RaaS), so capable of existing ‘in the cloud’ – In this model, the ‘TTP as intermediary’ vs. ‘we don’t do business through TTPs’ tension is transformed. – Tracking risk, reputation and the satisfaction of obligations goes ‘into the cloud’ – TTP (now a ‘RaaS provider’) does not transfer or absorb counterparty risk/ liability Benefits • Today, risk/reputation scores, audit trails etc., are: – – – – After the fact, low-level, un-normalized Duplicated across enterprise architectures Bandwidth consuming if transmitted Subvertable • In contrast, the repository model seeks to: – – – – Unify this metadata In a normalized fashion Suitable for scalable multiparty access/update With denormalization required only • a) as by-product of implementation constraints, or • b) where one party needs to place greater trust in a local copy of repository data than another. Implications • Memetic view of collaboration allows collaborations to become ‘first-class citizens’: informs business architecture • Viewed ‘memetically’: – A transaction is the vehicle for a COntrAct replicator – A business outcome (potentially, risk impact) is the vehicle for a REPutation replicator • This requires changes in both the ‘business mindset’ and architectural assumptions about: – Where to put security metadata e.g. ‘classifications’ sit within COntrActs as a view of risk appetite – Relationships between security metadata, other metadata and transactional information flows ‘Bottom Line’ • We postulate that – appropriate collaborative team OD, design of incentives for more-or-less altruistically motivated team members, and – architectural underpinning for team working (using COA) • …together maximise effective, e-enabled collaboration. – This requires validation. – COA foundations (inherently secure communications, endpoint security etc.,) are necessary building blocks before RaaS can be reliably implemented. Next Steps • ‘Managing Collaboration Effectively’ open discussion group – This Summer, probably at London Business School (LBS) - expressions of interest sought – OD implications of guild models – Collaboration as a tool for ‘wicked’ problemsolving – Collaboration effectiveness • COA development – Join Jericho Forum to participate! Q&A • Thanks! • Contact: [email protected]