UbiCom Book Slides Chapter 9: Intelligent Interaction (All Parts, Short Version) Stefan Poslad http://www.eecs.qmul.ac.uk/people/stefan/ubicom Ubiquitous computing: smart devices, environments and interaction 1 Chapter 9: Overview Chapter 9 focuses on: • Internal system properties: intelligence • External interaction with any of three types of environment – Focussing more on ICT and physical environment – These environments may be active, i.e., they are themselves one or more intelligence systems Ubiquitous computing: smart devices, environments and interaction 2 Five main properties for UbiCom Handling Non-determinism Knowledge & task sharing Goal-based, etc. Ubiquitous computing: smart devices, environments and interaction 3 Ubiquitous computing: smart devices, environments and interaction 4 Related Chapter Links • There are two AI chapters that are interlinked • Chapter 8, describes the design of single Intelligent System or IS – These may be simple: use a single models of intelligence – These may be hybrid: use multiple heterogeneous intelligence models • This Chapter 9, describes intelligent interaction between multiple systems – The systems interacting may be intelligent (Chapter 8) – The systems interacting may not necessarily be intelligent but their interaction may still be: emergent intelligence (Chapter 10) – Or Both Ubiquitous computing: smart devices, environments and interaction 5 Related Chapter Links • Other Models of Interaction Multiplicity: – Service Interaction (Chapter 3) – Network Interaction (Chapter 11) • Each type of smart device to smart environment (device) interaction can be enhanced by making them intelligent – CCI (Chapters 3 & 4) – CPI (Chapters 6 & 7) – HCI (chapter 5) Ubiquitous computing: smart devices, environments and interaction 6 Chapter 9: Overview The slides for this chapter are also expanded and split into several parts • Part A: Interaction Multiplicity: Between peers • Part B: Interaction Multiplicity: Using Mediators • Part C: Cooperative Interaction • Part D: Competitive Interaction • Part E: Intelligent Interaction Protocols 1 • Part F: Intelligent Interaction Protocols 2 • Part G: Multi-Agent Systems • Part H: Social Interaction Ubiquitous computing: smart devices, environments and interaction 7 Part A Overview • Basic Smart versus Intelligent Interaction? • Interaction Multiplicity • P2P Interaction between Multiple Senders and Receivers • Unknown Sender and Malicious Senders • Unknown Receivers • Too Many Messages Ubiquitous computing: smart devices, environments and interaction 9 Introduction • Deployment of UbiCom is ? • UbiCom device interaction ? Ubiquitous computing: smart devices, environments and interaction 10 Basic (Smart) Interaction • P2P Interaction • Interaction that involves passive intermediaries Ubiquitous computing: smart devices, environments and interaction 11 Intelligent (Smart) interaction What is Intelligent (Smart) interaction? • Beyond using universal network communication protocols, • Involves Coordination • Use of Semantics • Communicate using a rich language • Organisational interaction • etc Intelligent interaction is built upon basic interaction Ubiquitous computing: smart devices, environments and interaction 12 Intelligent Interaction 2 dimensions of intelligent interaction • Interaction between multiple intelligent systems & their environments • Intelligent Interaction between relatively non-intelligent multiple systems & environments Ubiquitous computing: smart devices, environments and interaction 13 Smart Device versus Intelligent Device Interaction Interaction between smart devices (Chapter 1): • Digital, • Connected, • Degree of local autonomous control, etc Interaction between intelligent devices (Chapter 8): • Specific notions of intelligence, – e.g., reflexive, goal-based etc • Different degrees of intelligence (Chapter 13) Ubiquitous computing: smart devices, environments and interaction 14 Intelligent Environment versus Intelligent Interaction Intelligent Environment: • Environment for a system, is intelligent • Environment may include other intelligent systems Intelligent Interaction: • Interaction, between a system and its environment (including other systems), is intelligent Ubiquitous computing: smart devices, environments and interaction 15 Interaction Multiplicity • Interaction multiplicity can occur in many different components of UbiCom systems & their environments, e.g., – ICT Environment (C) • Services (S) • Networks (N). – Human Environment (H) – Physical Environment (P) Ubiquitous computing: smart devices, environments and interaction 16 Basic Interaction Multiplicity Example: Service Invocation Types of Interaction Multiplicity Interaction Multiplicity • Interaction multiplicity complexity of interaction. Why? • How to manage complexity of Interaction multiplicity? – . Ubiquitous computing: smart devices, environments and interaction 19 Interaction Multiplicity Examples: Communication Chapter 9: Overview The slides for this chapter are also expanded and split into several parts • Part A: Interaction Multiplicity: Between peers • Part B: Interaction Multiplicity: Using Mediators • Part C: Cooperative Interaction • Part D: Competitive Interaction • Part E: Intelligent Interaction Protocols 1 • Part F: Intelligent Interaction Protocols 2 • Part G: Multi-Agent Systems • Part H: Social Interaction Ubiquitous computing: smart devices, environments and interaction 21 Part B Overview • • • • Interaction using Mediators Shared Communication Resource Access Shared Computation Resource Access Mediating Between Requesters and Providers Ubiquitous computing: smart devices, environments and interaction 22 Mediated Interaction • Mediator: a go between interacting participants / peers – also referred to as 3rd parties, intermediaries, middle- ware -agents • Benefits: – Enhances peer discovery and service discovery – Etc • Disadvantages: – Performance drops as extra intermediate nodes / hops are used – etc Ubiquitous computing: smart devices, environments and interaction 23 Mediated Interaction • The following types of mediated interaction: – – – – Shared Communication Channel Access Shared Computation Resource Access Service Discovery etc • Can be considered in terms of: – Motivation? – Challenges? – Handled by? Ubiquitous computing: smart devices, environments and interaction 24 Mediating Between Requesters and Providers • Mediators enhance peer discovery & service discovery • Instead of having to request information from each peer, info. accessed in 1 place, a 3rd party at a well known, static, address uses well standardised directory interface • 2 types of information used in discovery process – service capability – Service preferences • Different designs for mediators exist depending on how service capabilities or preferences are kept private versus shared Ubiquitous computing: smart devices, environments and interaction 25 Who Knows Requests & Preferences Mediators Who Knows Capabilities Mediator Design Issues • When are Mediators used during an interaction? • Support for anonymity • Mediators can be designed to support a range of different representations for capabilities and preferences • Mediators can be designed to support different types of interaction • Mediator fairness to providers • Trust & Neutrality Ubiquitous computing: smart devices, environments and interaction 27 Mediators: Interaction Protocols • What interaction protocols should these different types of mediator use? – Request-reply? – Asynchronous notifications? – Other protocols? But how can we support richer & more flexible interaction? See later Ubiquitous computing: smart devices, environments and interaction 28 Chapter 9: Overview The slides for this chapter are also expanded and split into several parts • Part A: Interaction Multiplicity: Between peers • Part B: Interaction Multiplicity: Using Mediators • Part C: Cooperative Interaction • Part D: Competitive Interaction • Part E: Intelligent Interaction Protocols 1 • Part F: Intelligent Interaction Protocols 2 • Part G: Multi-Agent Systems • Part H: Social Interaction Ubiquitous computing: smart devices, environments and interaction 29 Part C Overview • • • • • • • Interaction using Cooperative Participants Coordination Basics Perfectly Coordinated Systems Coordination design issues Coordination through join intentions and plans Coordination using Norms and Electronic Institutions Hierarchical and Role-based Organisational Interaction Ubiquitous computing: smart devices, environments and interaction 30 Interaction Multiplicity: Cooperative Interaction • Cooperative interaction enables multiple systems to work together. Characterised by 2 main properties: • Coordination: synchronising activities • Cohesion: acting together (organisational interaction). Ubiquitous computing: smart devices, environments and interaction 31 Interaction Multiplicity: Cooperative Interaction • Cooperation is easier to manage when: – – – – – homogeneous designed systems interact; there is centralised control; systems are designed as pure servers systems are designed statically to cooperate; systems act benevolently and reliably. • Cooperation is harder to manage when: – – – – – – different systems are designed by independent developers; systems are designed to act autonomously; systems support heterogeneous goals; systems need to cooperate dynamically; parties may act in a self-interested manner; systems act malevolently and may non-deterministically malfunction. Ubiquitous computing: smart devices, environments and interaction 32 Cooperative Interaction: Pros and Cons Advantages of Cooperative Model Disadvantages of cooperative model Distributed problem solving: solves it Cooperation reduces and quicker as more parts are processed in competition can parallel. Delegation: Don’t need to do everything Communication, costs, unreliability ourselves. Don’t want to do it ourselves, may outweigh the extra processing too time-consuming. Instead delegate benefits of the distribution. Selection: select best option from a set of Coordination & management is candidates more complex: disruptions (insider attack), lack of understanding, ambiguity, conflict. Reliability: there are alternative options. Delegation and session initiation costs too high Social: agents act on behalf, to engage Lack of control, privacy. people Ubiquitous computing: smart devices, environments and interaction 33 Cooperative Interaction: coordination • Explicit Coordinated Cooperation • Coordination using Norms and Electronic Institutions • Hierarchical and Role-based Organisational Interaction Ubiquitous computing: smart devices, environments and interaction 34 Coordination: Classification Message-based vs. process-based Explicit vs. Implicit Perfect vs. imperfect etc Ubiquitous computing: smart devices, environments and interaction 35 Coordination Design Issues • Whether or not ISs are spatially and or temporally coincident, or not • Handling inconsistencies and uncertainty. How? – • Who Coordinates? – Ubiquitous computing: smart devices, environments and interaction 36 Coordination Classification Explicit coordination • Service composition(Section 3.3.4). • Interaction protocols with inbuilt coordination mechanisms – See later • Joint planning • Joint intentions Implicit coordination • Norms and Electronic Institutions • Hierarchical & Role-based Organisational Interaction Ubiquitous computing: smart devices, environments and interaction 37 UbiCom System Applications of Social Organisations Provide a flexible way to design a large range of organisations • to dynamically configure building facilities to support building energy efficiency; • for personalised work environments • for information integration and interoperability • information services for mobile users in which IS dynamically adapt information to multiple contexts such as location, person and ICT system. Ubiquitous computing: smart devices, environments and interaction 38 Chapter 9: Overview The slides for this chapter are also expanded and split into several parts • Part A: Interaction Multiplicity: Between peers • Part B: Interaction Multiplicity: Using Mediators • Part C: Cooperative Interaction • Part D: Competitive Interaction • Part E: Intelligent Interaction Protocols 1 • Part F: Intelligent Interaction Protocols 2 • Part G: Multi-Agent Systems • Part H: Social Interaction Ubiquitous computing: smart devices, environments and interaction 39 Part D Overview • • • • Interaction with Self-interested Participants Market-based Interaction and Auctions Negotiation and Agreements Consensus-based Agreements Ubiquitous computing: smart devices, environments and interaction 40 Competitive Interaction • Cooperation vs. Competition – cooperators share their goals with collaborating parties & act together – competitors keep their goals private & act self-interestedly.to further their own goals, rather on collaborating to help further others’ goals • As diverse smart autonomous, configurable, networked devices in physical spaces, competitive interaction • Design models to solve the associated resource conflicts and resource allocation problems will become essential. • E.g., in Smart utility regulation scenario. – Multiple autonomous lighting devices in smart environment, all seek to switch themselves on but some are redundant & wastes energy. – Multiple users may seek to configure a shared or multiple devices that overlap in function in multiple ways, e.g., multiple users wish to regulate heating and lighting levels differently. Ubiquitous computing: smart devices, environments and interaction 41 Competitive Interaction • Different types of competitive interaction problems and designs depending on: – – – – No. of players interaction protocols; Strategies; Nature of the completion • Self-interested interaction is complicated further when participants act maliciously, i.e., lie and collude. Ubiquitous computing: smart devices, environments and interaction 42 Interaction Multiplicity: Competitive Interaction Types • Market-based Interaction and Auctions: used to allocate resources to individual requesters • Negotiation and Agreements: more general than auctions, used in market places to agree terms but can also be more generally used to resolve conflicts. • Convergence: a multi-step processes where two or more entities iteratively reach an agreement. Convergence algorithms and protocols tend to be domain specific • Consensus based protocols can be used to reach agreement between multiple participants, but normally for one object at one time, e.g., voting protocols Ubiquitous computing: smart devices, environments and interaction 43 Competitive Interaction: Designs • A generic problem for UbiCom is allocation of limited resources & services to multiple self-interested requestors. Designs to manage this? • Control can be more generally acceded to a third party • Concurrency control (Section 9.2.2.2). • Policy based management (Section 12.2.8.4) • Market-based Interaction and Auctions • Negotiation and Agreements • Consensus-based Agreements • Can we classify these designs into types of mediator (passive versus active) & mediated versus non-mediated? Ubiquitous computing: smart devices, environments and interaction 44 Auctions • Is 1 of oldest but still widely-used market based protocols • Designed to allocate resources such as goods and services to one of the bidders. • Several types of auction protocol depending on – ???? • Auction Benefits – ??? Ubiquitous computing: smart devices, environments and interaction 45 Auctions • Different types of auction? • English auction, can be classified in terms of the following properties for bids: – – – – a single type of goods single attribute single sided ascending Ubiquitous computing: smart devices, environments and interaction 46 Negotiation & Agreements • Auctions are designed to reach agreements between sellers and consumers in a market-place – are considered to be a type of more general technique called negotiation. • General aims of Negotiation is modification of local agent policies to constrain interaction and plans of interaction, – e.g., in the case of negative (harmful) interactions, and identification of situations where new potential interactions are possible and beneficial. • Uses of negotiation in UbiCom? – – – – task and resource allocation; recognition of conflicts; resolution of goal disparities; determination of the organisational structure and hence for organisational coherence. Ubiquitous computing: smart devices, environments and interaction 47 Negotiation Design In general a negotiation method has 4 principle components: • Public shared interaction protocol • Deal rule • Negotiation set • Strategies that are kept private Design properties for negotiation protocols ? • pareto optimal, • stable • individually rational • support computation and communication efficiency Ubiquitous computing: smart devices, environments and interaction 48 Negotiation Designs • Negotiation can be considered to be a distributed search – Search is through a space of potential agreements (chapter 8) • Game-theory is used to develop strategies between competing players who strive to win a game • Argumentation-based negotiation allows additional information to be exchanged, over and above proposals • Different problem domain models for negotiation applications: – Task-based – State-based – Worth-based. Ubiquitous computing: smart devices, environments and interaction 49 Consensus-based Agreements • Consensus based interaction can be used to reach an agreement when multiple self-interested participants share a common goal • Consensus is important when different participants or processes interact such that their self-interested goals may conflict • Consensus refers both to a state of agreement that is reached by independent participants and to the process to reach an agreement. • Consensus may also be useful in situations where there are several alternatives but it is not clear which one alternative should be chosen, • In contrast to negotiation, consensus based agreements are simpler. How? Ubiquitous computing: smart devices, environments and interaction 50 Chapter 9: Overview The slides for this chapter are also expanded and split into several parts • Part A: Interaction Multiplicity: Between peers • Part B: Interaction Multiplicity: Using Mediators • Part C: Cooperative Interaction • Part D: Competitive Interaction • Part E: Intelligent Interaction Protocols 1 • Part F: Intelligent Interaction Protocols 2 • Part G: Multi-Agent Systems • Part H: Social Interaction Ubiquitous computing: smart devices, environments and interaction 51 Part E Overview • IS Interaction Design • Designing System Interaction to be more Intelligent • Designing Interaction between Individual Intelligent Systems • Interaction Protocol Design Ubiquitous computing: smart devices, environments and interaction 52 IS Interaction Design 2 basic dimensions to supporting intelligent interaction: • to design conventional system interaction to be intelligent • to design individual intelligent systems to interact Ubiquitous computing: smart devices, environments and interaction 53 More Intelligent Conventional system Interaction Motivation • Mediation & handling heterogeneity • Reflection about communication • Distributed problem solving • Task delegation • Flexibility and Selection • Reliability Ubiquitous computing: smart devices, environments and interaction 54 Interaction Between Individual ISs: Motivation • Part of the motivation for individual intelligent systems to interact with each other, is to handle the knowledge bootstrapping problem. How? • Single intelligent entity needs to independently learn everything it needs to know itself • Single intelligent entity would also need its own internal, complete, knowledge model of the world and of itself, Ubiquitous computing: smart devices, environments and interaction 55 IS Interaction: Design Issues Interaction design issues: • If common, extensible message protocol can be designed for use across multiple types of UbiCom interaction • • If ISs need to share and fix a common understanding of terms or concepts within a domain • If ISs need to share the some context associated with a message, Ubiquitous computing: smart devices, environments and interaction 56 IS Communication Protocols • Involves specifying 2 separate (sub-) application layer protocols: – specifying individual messages (Message Protocols) – Specifying patterns of multiple messages (Interaction Protocols) • Message protocols define – ???? • Interaction protocols define – ???? Ubiquitous computing: smart devices, environments and interaction 57 Interaction Protocol Types • Individual messages are not used in isolation but used in different patterns of multiple messages Classification of Interaction Protocols? • Information sharing vs. task sharing • Unicast versus Multicast • Pull versus Push • Syntactic versus Semantic versus Linguistic Ubiquitous computing: smart devices, environments and interaction 58 Handling Interaction Failures How can the following interaction failures be handled? • Network link failure • Receiver down, not ready • Wrong message syntax • Use wrong default values, types • Use of service constraints that cannot be satisfied by provider • Unknown service providers and location • Client action, e.g., sender cancel • Messages as part of processes not coordinated • Semantic differences in use of terms at sender & receiver Ubiquitous computing: smart devices, environments and interaction 59 Interaction Protocol Design Issues • Interaction protocols are often designed to be servicespecific and domain specific – -> service interoperability is challenging • Interaction protocols are fixed, not very extensible – Although multiple interaction protocols can be orchestrated • Can introduce interaction flexibility – through use of cooperative dialogues – Explicitly supporting, nesting, concurrency etc • Can add interaction richness – Through use of semantic protocols to define meaning, context of interaction – But how to define the semantics – Using, OWL, Speech Acts? Ubiquitous computing: smart devices, environments and interaction 60 Chapter 9: Overview The slides for this chapter are also expanded and split into several parts • Part A: Interaction Multiplicity: Between peers • Part B: Interaction Multiplicity: Using Mediators • Part C: Cooperative Interaction • Part D: Competitive Interaction • Part E: Intelligent Interaction Protocols 1 • Part F: Intelligent Interaction Protocols 2 • Part G: Multi-Agent Systems • Part H: Social Interaction Ubiquitous computing: smart devices, environments and interaction 61 Part F Overview • Semantic and Knowledge Sharing Protocols • ACLs and Linguistic-based Protocols • Examples of use of Interaction Protocols in PM Scenario Ubiquitous computing: smart devices, environments and interaction 62 Semantics of Communication Protocols • Communication protocols (CP) are specified in terms of human-readable but not in machine-readable semantics. • Lack of a standard semantic representation for protocols Can standard KRs for content, e.g., OWL, be used for CP? Ubiquitous computing: smart devices, environments and interaction 63 IS Message Protocols Based Upon Speech Acts • Based upon type of linguistic protocol called Speech Acts • Some speech utterances are like physical actions that change the state of the world, – e.g., pronouncing someone as ‘man & wife’ in a religious ceremony, – sending message to set a new fact that changes the state of KB. • Basic structure of speech act defines: – Type of action – Pre-conditions which if true enable actions to be triggered. – Post-conditions of effects define what should now be true if the action was successfully executed. • The most useful types of communicative acts are – Assertives: set (information, facts, system states) – Directives: task requests, info. queries, mediating actions – Phatics: establish, check, prolong and interrupt, control comms Ubiquitous computing: smart devices, environments and interaction 64 Speech Act Model Benefits • Why used be speech act-based interaction between ISs? – Rather than conventional communication protocols between Iss • Generic model of communicative acts could be used across all knowledge domains, enhancing service interoperability. • In contrast, currently, each application domain and even multiple applications within that domain specify their own sets of service actions. – This makes interoperability using service actions defined in heterogeneous service models complex. • Some instantiation of service actions is needed to ground the semantics – this could vary across services and service domains. Ubiquitous computing: smart devices, environments and interaction 65 Speech Act Semantic specifications for ACLs: • BDI • Contract programming model semantics • Semantic Commitments based upon social conventions • IP context can be used as the semantics for the communicative act. – E.g., FIPA Interaction Protocol model makes a rudimentary attempt at a social model in the sense that the interaction is related to the organisational roles of the interacting parties and the semantics of each CA in an IP is interpreted within the context of the IP. Ubiquitous computing: smart devices, environments and interaction 66 Interaction Protocol Example: Request • Specifications of communication using speech acts – how multiple communicative acts can be used as part of different interaction patterns – how a communicative act links to the message content • Request interaction pattern – Classification: a task-sharing one-to-one pull type interaction • Although it seems similar to a client-server type requestresponse pattern, it is more flexible in several ways – responder can optionally choose to acknowledge request & supply the result later. – responder also has different options to signal: • lack of understanding of the request, • failure for some reason such as lack of sufficient ICT resources • refusal where although it could do the task for some reason, it chooses not to. Ubiquitous computing: smart devices, environments and interaction 67 Interaction Protocol Example: Request ELEM007 Agents 68 Interaction Protocol Use in PM Scenario • As an example of the benefits of using rich and flexible interaction protocols, resource access is considered during personal memories (PM) scenario – e.g., accessing and displaying or playing audio-video content. • In the simple case, we invoke an AV-player (service) passing the details of the source of the AV content of our choice. • But when we play the AV source it fails, the system must then decide how to proceed? Ubiquitous computing: smart devices, environments and interaction 69 Interaction Protocol Use in PM Scenario • Typically, requester asks for assistance by searching for help in a well-known place a directory. • Once the requester finds a help assistance, the requester can choose to delegate the resource access task to the help assistant – providing some conditions were fulfilled such as authentication and competency checks were fulfilled Ubiquitous computing: smart devices, environments and interaction 70 Interaction Protocol Use in PM Scenario In more detail, • A Help peer registers itself with provider to be informed when its fail to fulfil a request from a requester (a subscribe). • The Help peer then announces itself to the requester. The requester then queries Help about use of a resource X. • Help advices A to ask resource depository D something. • D tells the requester that peers E,F,G, H have resource X. • Requester issues a contract (a call for proposals or cfp) to E,F,G,H to ask them to bid to supply the resource as A does not know which of these can provide the resource most favourably • etc. Ubiquitous computing: smart devices, environments and interaction 71 Interaction Protocol Use in PM Scenario Resource Requester Resource Repository Help Assistant request query inform inform query inform inform Delegating task of resource access to a help assistant Chapter 9: Overview The slides for this chapter are also expanded and split into several parts • Part A: Interaction Multiplicity: Between peers • Part B: Interaction Multiplicity: Using Mediators • Part C: Cooperative Interaction • Part D: Competitive Interaction • Part E: Intelligent Interaction Protocols 1 • Part F: Intelligent Interaction Protocols 2 • Part G: Multi-Agent Systems • Part H: Social Interaction Ubiquitous computing: smart devices, environments and interaction 73 Part G Overview • • • • Multi-Agent Systems ACL and Agent Platform Design Multi-Agent System Application Design Some Generic Intelligent Interaction Applications Ubiquitous computing: smart devices, environments and interaction 74 Multi-Agent Systems (MAS) • Generally, if an IS is represented by an agent, then a MAS represents multiple interacting IS. • When MAS interact with other MAS they represent systems of systems interacting. • Can characterise the properties of MAS? – – – – – • degree of dynamism degree of scale (numbers of agents) type of (organisational) control homogeneous versus heterogeneous types of individual agent type of agent interaction (e.g., goal exchange, belief exchange etc). Use of an appropriate ACL (Agent Communication Language) can support these MAS properties. Ubiquitous computing: smart devices, environments and interaction 75 Agent Platform or Middleware Design • Common Multi-Agent Systems (MAS) consists of: – Agent Interaction Protocol Suite (AIPS): individual agents interact using an Agent Communication Language – Agent Platform or Middleware accessed through some API – MAS Applications Ubiquitous computing: smart devices, environments and interaction 76 Agent Platform or Middleware Design • Core agent middleware services typically include: – ACL interaction – agent name / agent life-cycle management – directory facilitator service • Should each service be an agent? • Several agent toolkits have been developed which support the FIPA ACL and agent platforms , e.g., JADE Ubiquitous computing: smart devices, environments and interaction 77 Agent Interaction Protocol Suite • In terms of the TCP/IP protocol suite, • ACL behaves as a suite of multi-layer protocols at the application level – for this reason has been termed an AIPS or Agent Interaction Protocol Suite or AIPS • Several protocols are needed to support interaction for intelligent applications using communicative acts: – – – – an interaction protocol a communicative act protocols a content protocol. Content protocol may separate the content Ontology from a content logic. Why? Ubiquitous computing: smart devices, environments and interaction 78 Agent Interaction Protocol Suite MAS Application Design • Many examples of IS or Agent-Oriented Software Engineering (AOSE) methodologies (also called AgentOriented Development or AOD). • There are 2 basic types: – those which extend or adapting non-Is system methodologies, e.g., object-oriented based AOSE – those based upon AI methodologies. – Hybrid methods Ubiquitous computing: smart devices, environments and interaction 80 MAS Application Design • AOSE design can be captured in two main model views: – organisational view: specifies types of agents & roles – operational view: specifies interaction constrained by goals and plans of actions to achieve those goals. • (Organisational) Roles support a more dynamic approach. • Roles used to separate responsibility for service access from identity to: – enable agents to combine multiple roles; – enable several agents to play same role (redundancy); – enable agents to change roles at run time; • Plans: order system (internal) actions & (external) interactions – interactions that determine the roles of entities in the organisation • . Ubiquitous computing: smart devices, environments and interaction 81 Organisational Model for Part of PM Scenario Operational Model for Part of PM Scenario Operational Model for Part of PM Scenario • Example of an operational view is a simple task-oriented description of a problem • Goal is to display images from the digital camera peer on some (visual) display, i.e., part of the personal memories scenario. • Goal is normally achieved using a very simple default plan which in this case consists of a sequence of two actions: – AV source peer such as a digital camera or its storage media requests use of image reading functions of an AV player which is connected to a display – Provider then responds by signalling to the AV source that the display is ready. Ubiquitous computing: smart devices, environments and interaction 84 Chapter 9: Overview The slides for this chapter are also expanded and split into several parts • Part A: Interaction Multiplicity: Between peers • Part B: Interaction Multiplicity: Using Mediators • Part C: Cooperative Interaction • Part D: Competitive Interaction • Part E: Intelligent Interaction Protocols 1 • Part F: Intelligent Interaction Protocols 2 • Part G: Multi-Agent Systems • Part H: Social Interaction Ubiquitous computing: smart devices, environments and interaction 85 Part H Overview • Social Networking & Media Exchange • Recommender Systems – Content-Based – Collaborative Filtering • Referral Systems • Pervasive Work Flow Management for People • Trust Management Ubiquitous computing: smart devices, environments and interaction 86 Introduction • There are many intelligent interaction applications which can be classified into: – CCI, e.g., autonomous systems; – HCI – CPI. • This section focuses on social type HCI type interaction models and their application to CCI and CPI. Ubiquitous computing: smart devices, environments and interaction 87 Social Networking and Media Exchange • Humans have an instinctive need to communicate, to socialise • Providing content for public and private access is now cheap relative to the average standard of living. • Remote Social interaction Ubiquitous computing: smart devices, environments and interaction 88 Social Networking and Media Exchange: Challenges • Remote Social interaction but at the expense of Local social interaction • Inclusion: requires ICT resources not affordable for everyone, everywhere? • New means of delivering content leaves us open to the potential hazards that exist in the physical world? Ubiquitous computing: smart devices, environments and interaction 89 Social Networking and Media Exchange: Semantics • Semantics used to enable semantic interoperability,. • This misuse can potentially benefit from the semantics. Why? Ubiquitous computing: smart devices, environments and interaction 90 Social Media • Social Media experience can take many different forms, including text, images, audio, and video. • Popular mediums include: – – – – – Blogs & Microbloggs, e.g., twitter Social networks, Content communities (sometimes called folksonomies) podcasts wikis, Ubiquitous computing: smart devices, environments and interaction 91 Social Media: Organisation • In general the trend is that large amounts of content is created and shared by users and a stronger move to Web as a user-driven application platform. • It is not feasible to expect even the diligent users to annotate and add details to their content and organise this in a methodical and consistent. • To help the users to manage and organize their content requires certain contextual knowledge that comes from a number of places – e.g. content annotations, semantic metadata, contact lists, the way the user organises contact lists as family and friends, etc. Ubiquitous computing: smart devices, environments and interaction 92 Social Media: Organisation • To organize content for users requires that a system has certain pertinent and significant knowledge about users, their context & habits. • Creation of contextual knowledge & use of contextual knowledge can help user to manage & share their content • Challenge in any solution is in the way the knowledge is aggregated to provide a contextual filter that can be applied to the organization of the content. Ubiquitous computing: smart devices, environments and interaction 93 Social Media: Organisation There are different techniques to filter and organise media • Personalisation • Self-organisation • Self-governance . Ubiquitous computing: smart devices, environments and interaction 94 Recommender Systems • Recommenders are types of personalisation software • Is often viewed as crucial for e-commerce sites. • There are two main types of recommender system: – Content-based filtering – Collaborative-filtering Ubiquitous computing: smart devices, environments and interaction 95 Recommender Systems • Examples of the use of recommender system for Ubicom? Ubiquitous computing: smart devices, environments and interaction 96 Referrals • Referrals are trusted recommendations by known people, in contrast to recommendations that are anonymous. • For serious life and business decisions, people often value the opinion of a trusted expert more, rather than an anonymous decision. • Discuss how referral (chains) work here Ubiquitous computing: smart devices, environments and interaction 97 Pervasive Work Flow Management for People • Physical distance is much less of a barrier for communication and workflow • Virtual distance between employees in terms of differing beliefs, systems and experiences is still a barrier. • Effectiveness of the cooperation between them. Why? – Lack of trust between different layers of organisation • Solutions? Ubiquitous computing: smart devices, environments and interaction 98 Trust • Trust is an inherent property in UbiCom systems in which: – one autonomous component cannot completely control another autonomous component – but which may need to rely or one another or require some cooperation from it. • Trust in social organisations, is a general expectation, explicitly, evaluated, that one autonomous component, the truster, can rely on another autonomous component, the trustee, in order to share information, tasks, goals, etc, with them. Ubiquitous computing: smart devices, environments and interaction 99 Trust Design Issues • There are several dimensions or metrics to specify trust for use in open UbiCom systems? – personal trust or impersonal trust – the disposition of the truster to trust, which ranges from being averse to trust to being eager to trust – if distrust is modelled as the complete absence of trust (zero trust) or is considered in terms of negative metrics for trust Ubiquitous computing: smart devices, environments and interaction 100 Trust-based Management • Social control based upon trust is sometimes referred to as a soft security in contrast to hard security which is control based on encryption type algorithms. • Soft security is viewed as a more effective mechanism for security, in terms of robustness, scalability, and adaptability, in pervasive environments such as information-sharing communities that support inter-organisational interactions • The relationship or dependency of the truster on the trustee is referred as a trust relation. Ubiquitous computing: smart devices, environments and interaction 101 Social Trust Use in UbiCom • In many computer systems, although a notion of trust may be implied, such as the use of a trusted platform, or a trusted third party, there is often no explicit computation model of trust incorporated. • Trust is more useful issue for external interaction rather than internal interaction. • Internal interaction is often designed to use well-defined notions of control which can obviate the need for trust. • In external interaction between one autonomous system and its environment or another autonomous system, use of a centralised control mechanism is not possible by design. Ubiquitous computing: smart devices, environments and interaction 102 Social Trust Use in UbiCom • Sometimes there are concepts which are akin to trust used in distributed systems? – System may define a QoS for another peer to provide – Requester can examine the collective reputation of another peer before deciding to interact with it. • Peers are also often defined to be eager to trust, to blindly trust, This is referred to as an ad hoc trust model. – e.g., if the provider has a service description in well-known directory then the provider must be trustworthy, etc. • Of course the directory may offer no control or checks about whether or not malicious providers can offer services or malicious clients can search for services, – hence blind trust in the directory service can be associated with an unknown trust relation between one peer and another. Ubiquitous computing: smart devices, environments and interaction 103 Social Trust Models Use in UbiCom • Benefits? • Ways to incorporated trust in a computation form in UbiCom systems. – Authentication-based policy systems based upon PKI. – Authorisation-based policy systems such as SPKI, Ubiquitous computing: smart devices, environments and interaction 104 Social Trust Model Use in Intelligent Systems • Many applications of multi-agent-system models use collaborative type filtering mechanisms – E.g., based upon recommendations and reputations as well as policy-based MAS models to support impersonal trust. • 2 main aspects of design trust for Multiple IS. – to allow agents to trust each other -> need to endow them with the ability to reason about reciprocative nature, reliability, or honesty of their counterparts. – to design protocols and mechanisms of interactions such that participants will find no better option than telling the truth and interacting honestly with each other. Ubiquitous computing: smart devices, environments and interaction 105 Summary & Revision For each chapter • See book web-site for chapter summaries, references, resources etc. • Identify new terms & concepts • Apply new terms and concepts: define, use in old and new situations & problems • Debate problems, challenges and solutions • See Chapter exercises on web-site Ubiquitous computing: smart devices, environments and interaction 106 Exercises: Define New Concepts Ubiquitous computing: smart devices, environments and interaction 107 Exercise: Applying New Concepts Ubiquitous computing: smart devices, environments and interaction 108