Communities of Network Agents for Electronic Commerce

1
Communities of Network Agents for Electronic
Commerce

• Michael Weiss, Carleton University
• weiss_at_scs.carleton.ca
• www.scs.carleton.ca/weiss

2
Introduction

• Pervasiveness of the Internet has enabled the
  creation of network communities
• Commonly accepted definitions of community
• A community is a social network of
  interactingindividuals, usually concentrated
  into a defined territory (Dictionary of Human
  Geography)
• A community is a body of individuals organized
  into a unit or manifesting, usually with
  awareness, some unifying trait (Webster)
• The factors of community include thus locality,
  social interaction and common tie

3
Issues

• Presence
• who is currently there?
• Awareness
• where to look for similar users?
• History
• when did an event happen?
• Personalization
• how to learn users' preferences?

• Negotiation
• how to enforce house rules?
• Privacy
• how to protect users from prying eyes?
• Trust
• how to create trust among strangers?
• Security
• how to exchange data in safe ways

4
Case in Point Trust

• A quote from F. Fukuyama, author of Trust The
  Social Virtues The Creation of Prosperity,
  illustrates the role of social networks in
  building trust among members of a community
• Trust is not just about bits or bytes. Its
  about social relationships, and about building
  networks that deliver what they promise, be it a
  product, a collaboration, or simply reliable
  information.

Source http//www.ml.com/woml/forum/ecommerce1.ht
m
5
Community Applications

• A broad range of people can benefit from
  community support systems groups in industry,
  schools, universities, cities, hospitals, etc.
• The following scenarios illustrate some typical
  applications
• Virtual malls
• Customer support
• e-Marketplaces
• Ad hoc groups
• Network games
• Distance learning

6
Network Communities

• Informal group of actors formed over a network,
  formed around a set of common interests
• Participants are widely distributed (a
  centralized solution doesn't scale)
• Communities change their active membership
  changes, in addition to the roles and objectives
  of individuals and the community
• They preserve the individuality of their members
  participants can have diverse objectives

7
Purpose of Network Communities

• Purpose for forming a network community
• Sharing knowledge
• Linking members of the community
• Community support systems needs to provide shared
  contexts for actors to interact
• An example is a virtual meeting room actors are
  represented by avatars that signal their
  presence, and actors communicate through
  voice/text messages
• In order to interact with each other, actors need
  to locate each other (social network is tacit)
• who knows what?, who knows who?, who knows
  who knows who?, and who knows who knows what?

8
Virtual Meeting Room
Group
9
Virtual Meeting Room Architecture
World server
Client
Client
Client
10
Link to e-Commerce Systems

• e-Commerce systems demonstrate important
  characteristics of network communities
• Electronic marketplaces operate on a global scale
  
• Just as network communities are widely
  distributed
• It is dynamic in nature, as is a network
  community
• Traders are not known to each other in advance
• Demand and supply change continuously
• Each trader receives a personalized view on the
  marketplace, while preserving her privacy
• As network communities preserve an actors
  identity

11
Communities in e-Commerce

• Examples of network community processes at work
  in e-Commerce are
• Recommender systems (Guttman 99)
• Exchanges (Steinmetz 99)
• Reputation brokers (Zacharia 99)
• There is a common design pattern at the core of
  these examples
• Community members can be represented by a user
  agent and a number of role agents
• Community support infrastructure is provided by
  mediator agents (Voss 98, Hattori 99)

12
Community Design Pattern
Community
Personal unit
Mediator agent
User agent with its role agents
13
User Agent

• User agents form the interface between the user
  and the other agents
• They receive the user's queries and feedback, and
  present information tailored to the user
• The user agent manages the users profile and
  controls who can access it
• A user agent delegates the queries or orders
  (buys/sells) to a role agent
• It visualizes the structure of the community,
  which has been articulated via mediators

14
User Agent
instructs
User
Role
results
delegates
feedback
Buyer
Seller
15
Community Viewer
The CommunityViewer uses a spatial metaphor to
visualize the community. Users are located in 2D
space such that the distance among two users
reflects the degree of common interest among them
(as calculated from the users' profile data).
16
Agent As Mediator

• In a dynamic marketplace it is impossible for
  traders to maintain an up-do-date list of
  contacts
• Mediators are facilitators or infomediaries that
  set up relationships between trading parties
• Mediators provide shared information (e.g. user
  ratings of products), contexts for the
  interaction of role agents (e.g. market
  sessions), and facilities for introducing role
  agents to each other
• Mediators include directory services,
  transcoders, recommendation brokers, and market
  makers

17
Agent As Mediator
Role

Mediator
Wrapper
queries

consults
Buyer
Seller
18
Recommender Agent
19
Issue Information Overload

• People wish to find relevant information to make
  good deals and generate profit
• But there are too many providers of products and
  services and too many different interfaces
• One solution has been to provide portals or
  common entry points to the web
• These portals collect and condense information
  from a multitude of information sources into an
  index
• The disadvantage is that this index will be the
  same for every user (not based on their
  individual preferences)

20
Issue Ensuring Quality

• Shopping online lacks the immediate mechanisms
  for establishing trustworthiness
• How can you trust a vendor, with whom you had no
  previous encounter, whether the order you placed
  will be fulfilled satisfactorily?
• Some auction sites, e.g., solicit feedback about
  the performance of a vendor from customers
• eBay, for example, keeps records of how a vendor
  was rated by other customers
• Potential new customers will take this rating
  into account before considering buying from a
  vendor

21
Issue Search Costs

• It may be expensive for vendors and the customers
  to find each other
• In a static marketplace, each customer may store
  a contact list of vendors for each product
• However, the electronic marketplace is dynamic
  (vendors and customers come and go, their
  offerings and requirements change qualitatively
  and quantitatively)
• A mediator can match potential customers and
  vendors in the marketplace
• Reduces the need for customers and vendors to
  contact a large number of alternative partners

22
Issue Pricing

• It is generally inefficient for vendors to set a
  fixed price for their products or services
• Offline, information about customers is less
  readily available to vendors, and changing prices
  is often expensive
• Pricing is pushed into a marketplace, where
  prices are determined through negotiation
• As a result, limited resources are allocated to
  those who value them most
• However, negotiating may be too complicated or
  frustrating for the average consumer and
  consumers need to manage their negotiation
  strategies

23
e-Commerce Communities
Customer support
e-Commerce Community Applications
Recommender systems
Reputation brokers
Virtual malls
e-Marketplaces
Ad hoc groups
Direct Marketing
Expertise location
24
Customer Support

• Most typically one encounters customer support in
  the form of a call center
• Customers are guided through an Interactive Voice
  Response (IVR) system and then queued to wait for
  a service representative
• Service reps are statically grouped by criteria
  such as language, region and product calls are
  routed to service reps based on their grouping
  and availability
• Skills-based routing goes beyond todays call
  centers
• It matches the customer up with a service rep by
  comparing a problem profile to the service reps
  skill profiles

25
Skills-Based Routing
1. Select a rep
2. Establish call
3. Rate service
Call center
26
Social Appliances

• Another significant step towards network
  communities will be to employ mobile computing
  technologies (mobile agents and personal digital
  assistants) that would allow us to carry our
  access point to the network community with us
• There could be a broad range of such devices
• Meme-tags from MIT's ThingsThat Think project
  (Borovoy 98)
• Personalized navigation devices (Nagao 98)
• Devices to provide information at conferences
  (Nishibe 98)

27
Conclusion

• Many issues related to e-Commerce (e.g. trust
  establishment) dont require a technical
  solution, but need to be grounded in social
  relationships
• The social relationships between actors in an
  e-Marketplace are often only tacit and need to be
  articulated and visualized to tap into them
• Extracted a common design pattern from existing
  network community support systems and
  investigated how to apply it to address the
  issues identified for e-Commerce applications
  (such as trust, information overload, search
  costs, pricing)

28
Future Steps

• Document the patterns for building e-Commerce
  community applications
• Build an agent-based framework for articulating
  and visualizing community structure
• Investigate new types of e-Commerce applications
  enabled by network communities
• Direct marketing
• Ad hoc groups
• Expertise location
• Sales support

29
References

• Ishida, T. (ed.), Community Computing and Support
  Systems Social Interaction in Networked
  Communities, LNCS 1519, Springer, 1998
• Guttman, R, and Maes, P., Agents that Buy and
  Sell, Comm. ACM, 81-91, March 1999
• Hattori, F., Ohguru, T., et al, Socialware
  Multiagent Systems for Supporting Network
  Communities, Comm. ACM, 55-61, March 1999
• Noriega, P., and Sierra, C. (eds.),
  Agent-Mediated Electronic Commerce, LNAI 1571,
  Springer, 1999
• Steinmetz, E., Collins, J., et al, Bid Evaluation
  and Selection in the MAGNET Automated Contracting
  System, in (Noriega 99)
• Voss, A., and Kreifelts, T., SOAP Social Agents
  Providing People with Useful Information, in 5th
  DELOS Workshop on Filtering and Collaboarative
  Filtering, 1998
• Zacharia, G., Moukas, A., and Maes, P.
  Collaborative reputation mechanisms in electronic
  marketplaces. In Proceedings of the HICSS-99
  Conference, Electronic Commerce Minitrack, IEEE,
  1999