Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
1
Agent-Based Computing: Promise and Perils
Nick JenningsDept of Electronics and Computer Science
University of [email protected]
http://www.ecs.soton.ac.uk/~nrj/
2
Agent-Based Computing: A New Synthesis for AI and CS
Increasing number of systems viewed in terms of agents:
• as a theoretical model of computation n more closely reflects current computing reality than Turing
Machines
• as a model for engineering distributed software systemsn better suited than object-orientation, design patterns and
software architectures
• as a model for conceptualising and building intelligent entities n framework for unifying piecemeal specialisations
3
But fundamental questions remain:
• why are agents an appealing computational model?
• what are the drawbacks of an agent-oriented approach?
• what are the wider implications for AI and CS of agentbased computing?
• what is the essence of agent-based computing?
4
Which Perspective?
u Answer from many points of view• from philosophical to pragmatic
uProceed from standpoint of using agent-based software to solve complex, real-world problems
5
Software Development is Difficult♦ The most complex construction task humans undertake
“Computer science is the first engineering discipline ever in which the complexity of the objects created is limited by the skill of the creator and not limited by the strength of the raw materials. If steel beams were infinitely strong and couldn’t ever bend no matter what you did, then skyscrapers could be as complicated as computers.” Brian K. Reid
u True whatever models and techniques are applied:“the essential complexity of software” Fred Brooks
u Software engineering provides models &techniques that make it easier to handlethis essential complexity
6
The Continual Striving for Better Models and Techniques
♦ New approaches:
• component-ware
• design patterns
• application frameworks
• software architectures
• ……….
♦ Right direction, but:
• interactions too rigidly defined
• impoverished mechanisms for representing organisational context
7
The Adequacy Hypothesis
Agent-oriented approaches can significantly enhance our ability to model, design and build complex
(distributed) software systems
8
Exemplar Complex Systems
♦ Air traffic control
♦ Network management
♦ Manufacturing control
♦ Business process management
Complexity from large number of parts, many interconnections &
problem dynamics
Not algorithmic / computational complexity
9
Talk OutlineI. The Essence of Agent-Based Computing
II. Promise of Agent-Oriented Software Engineering
III. Perils of Agent-Oriented Software Engineering
IV. Conclusions
10
“The intolerable wrestle with words and meanings.”
T. S. Eliot
“We seldom attribute good sense, except with those who agree with us.”
Duc de la Rouchefoucauld
The Essence of Agent-Based Computing
11
“encapsulated computer system, situated in some environment, and capable of flexible autonomous action in that environment in order to meet its
design objectives” (Wooldridge)
Agent
12
“encapsulated computer system, situated in some environment, and capable of flexible autonomousautonomous action in that environment in order to meet its
design objectives” (Wooldridge)
Agent
♦ control over internal state and over own behaviour
13
“encapsulated computer system, situatedsituated in some environmentin some environment, and capable of flexible autonomous action in that environment in order to meet its
design objectives” (Wooldridge)
Agent
♦ control over internal state and over own behaviour
♦ experiences environment through sensors and acts through effectors
14
“encapsulated computer system, situated in some environment, and capable of flexibleflexible autonomous action in that environment in order to meet its
design objectives” (Wooldridge)
Agent
♦ reactive: respond in timely fashion to environmental change♦ proactive: act in anticipation of future goals
♦ control over internal state and over own behaviour
♦ experiences environment through sensors and acts through effectors
15
Definitional Malaise“My guess is that object-oriented programming will be what structured programming was in the 1970s. Everybody will be in favour of it. Every manufacturer will promote his product as supporting it. Every manager will pay lip service to it. Every programmer will practice it (differently). And no one will know just what it is.” (Rentsch, 82)
“My guess is that agent-based computing will be what object-oriented programming was in the 1980s. Everybody will be in favour of it. Every manufacturer will promote his product as supporting it. Every manager will pay lip service to it. Every programmer will practice it (differently). And no one will know just what it is.” (Jennings, 99)
16
Multiple Agents
In most cases, single agent is insufficient• no such thing as a single agent system (!?)
• multiple agents are the norm, to represent:(Bond & Gasser)
n natural decentralisationn multiple loci of controln multiple perspectivesn competing interests
17
Agent Interactions
♦ Interaction between agents is inevitable• to achieve individual objectives and to manage
inter-dependencies
♦ Conceptualised as taking place at knowledge-level • which goals, at what time, by whom, what for
♦ Flexible run-time initiation and responses• cf. design-time, hard-wired nature of extant approaches
paradigm shift from previous perceptions of computational interaction
18
CustomerServiceAgents
Provide Quote
Agent-Based Business Process Management (ADEPT)
Serviceproducer
Serviceconsumer
ServiceN
ame
(Jennings et al., 96)
19
CustomerServiceAgents
VetCustomer
Provide Quote
Agent-Based Business Process Management (ADEPT)
Customer Customer VettingVettingAgentsAgents
Serviceproducer
Serviceconsumer
A B C D n...ServiceN
ame
(Jennings et al., 96)
20
CustomerServiceAgents Legal
Agents
VetCustomer
Provide Quote ProvideLegalAdvice
Cost&DesignNetwork
Agent-Based Business Process Management (ADEPT)
Customer Customer VettingVettingAgentsAgents
Serviceproducer
Serviceconsumer
A B C D n...ServiceN
ame
DesignAgents
(Jennings et al., 96)
21
CustomerServiceAgents Legal
Agents
VetCustomer
Provide Quote ProvideLegalAdvice
SurveyPremisesCost&DesignNetwork
Agent-Based Business Process Management (ADEPT)
Customer Customer VettingVettingAgentsAgents
Serviceproducer
Serviceconsumer
A B C D n...ServiceN
ame
DesignAgents
SurveyAgents
(Jennings et al., 96)
22
In ADEPT, Interaction = Negotiation
A1 A2
XX X
X
Negotiation Space
ProposalsCounter-Proposals
AcceptsRejects
(Faratin, Sierra and Jennings, 98)
23
In ADEPT, Interaction = Negotiation
CostStart TimeEnd TimeQuality of ServicePenalty for Violation
Service Name
A1 A2
XX X
X
Negotiation Space
ProposalsCounter-Proposals
AcceptsRejects
(Faratin, Sierra and Jennings, 98)
24
In ADEPT, Interaction = Negotiation
CostStart TimeEnd TimeQuality of ServicePenalty for Violation
Service Name
A1 A2
XX X
X
Negotiation Space
ProposalsCounter-Proposals
AcceptsRejects
Does not know A2’s:
- reservation values- resource constraints- outside options
(Faratin, Sierra and Jennings, 98)
25
Why Negotiation?
♦ De facto means of coming to agreement between independent entities• cannot instruct an autonomous agent!
♦ Arrange inter-dependent activities (required services) according to prevailing situation:• short vs. long time to reach agreement• scarce vs. plentiful resource availability
26
Agents act/interact to achieve objectives:• on behalf of individuals/companies• part of a wider problem solving initiative
underlying organisational relationship between the agents
Organisations
27
Organisations in ADEPT
♦ Influence negotiation behaviour• Survey department accountable to design dept.• Peers in same organisation vs.
opponent in different organisation
♦Allow agents to be grouped• e.g. multiple designer agents working as a team
28
Organisations (Generally)
Since this organisational context:• influences agents’ behaviour
n relationships need to be made explicit– peers – teams, coalitions– authority relationships
• is subject to ongoing changen provide computational apparatus for creating,
maintaining and disbanding structures
29
A Canonical View
EnvironmentSphere of influence
AgentInteractions
Organisationalrelationships
(see also: Castelfranchi, Ferber, Gasser, Lesser, …..)
30
Talk OutlineI. The Essence of Agent-Based Computing
II. Promise of Agent-Oriented Software Engineering
III. Perils of Agent-Oriented Software Engineering
IV. Conclusions
ü
31
Making the Case: Quantitatively
0
10
20
30
40
50
Productivity Reliability Maintainability
Software Metrics
Objects Agents
“There are 3 kinds of lies: lies, damned lies and statistics”Disraeli
32
Making the Case: Qualitatively
33
Software techniques fortackling complexity
Making the Case: Qualitatively
34
(Grady Booch)
Tackling Complexity
♦Decomposition
♦Abstraction
♦Organisation
35
Nature of complexsystems
Software techniques fortackling complexity
Making the Case: Qualitatively
36
Complex Systems♦ Complexity takes form of “hierarchy”
•• notnot a control hierarchy• collection of related sub-systems at different levels of abstraction
♦ Can distinguish between interactions among sub-systems and interactions within sub-systems• latter more frequent & predictable: “nearly decomposable systems”
♦ Arbitrary choice about which components are primitive
♦ Systems that support evolutionary growth develop more quickly than those that do not : “stable intermediate forms”
(H. Simon)
37
Nature of complexsystems
Canonical view ofagent-based computing
Software techniques fortackling complexity
Making the Case: Qualitatively
38
Nature of complexsystems
Canonical view ofagent-based computing
Degree ofMatch
Software techniques fortackling complexity
Making the Case: Qualitatively
39
The Match Process
1. Show agent-oriented decomposition is effective way of partitioning problem space of complex system
2. Show key abstractions of agent-oriented mindsetare natural means of modelling complex systems
40
1. Decomposition: Agents
♦ In terms of entities that have:• own persistent thread of control (active: “say go”)
• control over their own destiny (autonomous: “say no”)
♦ Makes engineering of complex systems easier:• natural representation of multiple loci of control
n “real systems have no top” (Meyer)
• allows competing (sub-) objectives to be represented and reconciled in context sensitive fashion
41
Decomposition: Interactions
♦ Agents make decisions about nature & scope of interactions at run time
♦ Makes engineering of complex systems easier:• unexpected interaction is expected
n all interactions need not be set at design time
• simplified management of control relationships between componentsn coordination occurs on as-needed basis between
continuously active entities
42
2. Suitability of Abstractions
♦ Design is about having right models
♦ In software, minimise gap between units of analysis and constructs of solution paradigm• OO techniques natural way of modelling world
43
Complex System Agent-Based System
Sub-systems
Sub-system components
Interactions between sub-systems and
sub-system components
Relationships between sub-systems
and sub-system components
44
Complex System Agent-Based System
Sub-systems Agent organisations
Sub-system components
Interactions between sub-systems and
sub-system components
Relationships between sub-systems
and sub-system components
45
Complex System Agent-Based System
Sub-systems Agent organisations
Sub-system components Agents
Interactions between sub-systems and
sub-system components
Relationships between sub-systems
and sub-system components
46
Complex System Agent-Based System
Sub-systems Agent organisations
Sub-system components Agents
Interactions between sub-systems and
sub-system components:
“at any given level of abstraction, findmeaningful collections of entities that
collaborate to achieve some higher levelview” (Booch)
“cooperating to achieve common
objectives”
“coordinating their actions”
“negotiating to resolve conflicts”
Relationships between sub-systems
and sub-system components
47
Complex System Agent-Based System
Sub-systems Agent organisations
Sub-system components Agents
Interactions between sub-systems and
sub-system components
“cooperating to achieve commonobjectives”
“coordinating their actions”“negotiating to resolve conflicts”
Relationships between sub-systemsand sub-system components
- change over time- treat collections as single
coherent unit
Explicit mechanisms for representing &
managing organisational relationships
Structures for modelling collectives
48
üAgent-oriented approaches can
significantly enhance our ability to model, design and build complex
(distributed) software systems
The Adequacy Hypothesis
49
The Establishment Hypothesis
As well as being suitable for designing and building complex systems,
agents will succeed as a software engineering paradigm
[NB: will be complementary to existing software paradigms like OO, design patterns, …]
50
Agents Consistent with Trends in Software Engineering
♦ Conceptual basis rooted in problem domain• world contains autonomous entities that interact to get things done
51
Agents Consistent with Trends in Software Engineering
♦ Conceptual basis rooted in problem domain
♦ Increasing localisation and encapsulation• apply to control, as well as state and behaviour
52
Agents Consistent with Trends in Software Engineering
♦ Conceptual basis rooted in problem domain
♦ Increasing localisation and encapsulation
♦ Greater support for re-use of designs and programs• whole sub-system components (cf. design patterns, component-ware)
n e.g. agent architectures, system structures
• flexible Interactions (cf. application frameworks)n e.g. contract net protocol, auction protocols
• legacy softwaren e.g. place “wrapper” around existing code to give it agent interface
53
Agents Support System Development by Synthesis
An agent is a stable intermediate form• able to operate to achieve its objectives and
interact with others in flexible ways
construct “system” by bringing agents together and watching overall functionality emerge from their interplay
• well suited to developments in:n open systems (eg Internet)n e-commerce
54
Talk OutlineI. The Essence of Agent-Based Computing
II. Promise of Agent-Oriented Software Engineering
III. Perils of Agent-Oriented Software Engineering
IV. Conclusions
üü
“The moment we want to believe something, we suddenly see all thearguments for it, and become blind to the arguments against it”
George Bernard Shaw
55
Unpredictable Interactions
In general case:n patterns: who, what, and whyn outcomes: accept, reject, modify
decided at run-time by complex interplay of:
n agent’s internal staten environment’s (incl. acquaintances’) perceived staten organisational context
56
Emergent Behaviour
♦Behaviour cannot be understood solely in terms of individual components
“the whole is greater than the sum of the parts”
♦Can arise in any complex system• not always a bad thing
n sometimes a better model of problem
• more scope because of flexible nature of agent problem solving and interactions
57
Overcoming these Perils♦ Presently, can develop specific solutions for
specific problems• simple interaction protocols• rigid and pre-set organisational structures• limit nature and scope of agent interplay
♦ More generally, need better understanding of:
• impact of sociality and organisational responsibilities on individuals’ behaviour
• symbiotic relationship between individual and societal behaviour
58
Finding the Right Level
♦ Learn from GOFAI• Newell elucidated Knowledge Level (KL) as an abstraction
for addressing fundamental issues in single agent systems
♦ Propose Social LevelSocial Level (SL) (Jennings, 94; Jennings & Campos, 97)
• new computer level situated above KL• provides social principles and foundations for agent-based
systems
59
Dimension Description KL SL
System Entity to bedescribed
(asocial)Agent
Agent organisation
60
Dimension Description KL SL
SystemEntity to bedescribed
(asocial)Agent
Agent organisation
Components System’s primitiveelements
GoalsActions
Agents,Interaction channels,
Dependencies,Org. relationships
61
Dimension Description KL SL
System Entity to bedescribed
(asocial)Agent
Agent organisation
Components System’s primitiveelements
GoalsActions
Agents,Interaction channels,
Dependencies,Org. relationships
CompositionalLaw
How componentsare assembled Various
RolesOrganisation’s rules
62
Dimension Description KL SL
System Entity to bedescribed
(asocial)Agent
Agent organisation
Components System’s primitiveelements
GoalsActions
Agents,Interaction channels,
Dependencies,Org. relationships
CompositionalLaw
How componentsare assembled Various
RolesOrganisation’s rules
BehaviourLaw
How system’sbehaviour dependson composition &
components
Principle ofRationality
Principles of Organisational& Social Rationality
63
Dimension Description KL SL
System Entity to bedescribed
(asocial)Agent
Agent organisation
Components System’s primitiveelements
GoalsActions
Agents,Interaction channels,
Dependencies,Org. relationships
CompositionalLaw
How componentsare assembled
VariousRoles
Organisation’s rules
BehaviourLaw
How system’sbehaviour dependson composition &
components
Principle ofRationality
Principles of Organisational& Social Rationality
MediumElements processed
to obtain desiredbehaviour
Knowledge
Org. & Social obligationsMeans of influencing others
Means of changingorganisational structure
64
Exploiting the Social Level
♦Science of Agent-Based Computing• provides comprehensive model for specifying and
understanding behaviour in multiple agent systems
♦Engineering of Agent-Based Systems• provides basis for agent-based tools and
methodologies
65
Talk OutlineI. The Essence of Agent-Based Computing
II. Promise of Agent-Oriented Software Engineering
III. Perils of Agent-Oriented Software Engineering
IV. Conclusions
ü
üü
66
Promise
Agent-based computing has potential to:
provide powerful metaphors, concepts & techniques for conceptualising, designing and implementing complex (distributed) systems
67
Perils
♦ Insufficient understanding of:• impact of sociality on individual & collective behaviour• impact of organisational relationships on behaviour and
performance of individuals and societies
♦Remedied by:• distilling basis of computation in social settings
n Social Level characterisation is step towards this end
68
Acknowledgements
♦ Members of QMW DAI Group♦ Mike Wooldridge♦ Cristiano Castelfranchi♦ Ed Durfee♦ Les Gasser♦ Carl Hewitt
♦ Mike Huhns♦ Vic Lesser♦ Joerg MÜller♦ Simon Parsons♦ Van Parunak♦ Carles Sierra