Automated Negotiations

1
Automated Negotiations

• Yelena Yesha
• Olga Streltchenko

2
Automated Negotiations

• Issues in automated negotiations
• Auctions overview
• Electronic auctions (e-commerce servers)

3
Negotiation in electronic commerce

• Negotiation key component in e-commerce
• Two or more parties multilaterally bargain
  resources for intended gain, using the tools of
  electronic commerce,
• e.g. agents negotiating a solution
  electronically.
• Negotiating function is performed through
  (networked) computers.

4
Automated Negotiation

• Auto negotiation is performed by computational
  agents, which
• Represent real-world parties
• Perform information retrieval and processing
• Find prepare contracts
• Perform other activities .

5
Electronic marketplace

• Designated meeting place for negotiating
  parties.
• A trusted intermediary that facilitates trading
  between buyers and sellers on the Web.
• Closed'' marketplace
• predefined set of users
• enrollment.
• Open'' marketplace
• agents enter and exit any time.

6
Levels of automation

• Negotiation support systems.
• Help human negotiations.
• Intelligent Agents.
• Negotiate electronically within an environment
  governed by rules.
• No human intervention.

7
Automation

• Architecture Issues for Automated Markets
• Transaction processing.
• Decision support.

8
Architecture Issues for Automated Markets
Banking System

• E-Market entities (negotiating agents, etc.) have
  to be able to communicate with the existing
  banking and financial services.
• Integration of marketplace interfaces into
  banking legacy systems.
• Open standard for agent-to-bank communications.
• Security.

9
Architecture Issues for Automated Markets
Communication Infrastructure

• Efficiency and robustness.
• Redundancy
• E.g. a mesh of redundant hubs interconnected with
  each other.
• Open standard communication systems to allow
  development of platform-independent systems that
  plug into a marketplace architecture.
• Independent of agent architectures.
• Agents must
• access global posting services
• use common language for outbound communications.

10
Standardized Communication Infrastructure

• Integration into the back ends of electronic
  catalog databases and existing EDI systems.
• Common language for outbound communications
• KQML.

11
Architecture Issues for Automated Markets
Transfer and storage of goods

• Representation and handling of physical goods.
• Goods as software objects.
• Copy-protection to insure that an object is at
  one place at a time
• encoding by an owner
• access by agents authorized by the owner.
• Arrangement for physical shipment.

12
Architecture Issues for Automated Markets
Handling of Electronic Items

• Format of a software object.
• Copy-protection as above.
• Delivery channels.

13
Architecture Issues for Automated Markets
Administration and Policies

• Central administration to provide
• default protection
• prevention of illegal transactions
• collection of taxes and commissions
• credit and service ratings for agents.

14
Self-interested Agents (SI) vs Cooperative
(Distributed) Problem Solving (CDPM)

• Self-interested agents act according to their
  internal considerations and pursue their private
  goals
• E.g. internal utility function maximization.
• Competitive.
• A multi-agent system may strive to achieve a
  global (societal) goal
• E.g. global utility function maximization.
• Requires cooperation on the part of individual
  agents.
• Common goal in a distributed system.

15
Transaction Processing Levels of commitment

• For SI agents, contracts are bounding.
• In CDPS commitments are allowed to be broken
  unilaterally based on some local reasoning.
• Continuous levels of commitment based on a
  monetary penalty method.

16
Transaction Processing Decommitting

• Replies vs timeout.
• Inform the other party that the negotiation is
  not considered any more.

17
Transaction Processing Message congestion

• Most distributed implementations run into this
  problem.
• high risk of saturation.

18
Message Congestion (contd)

• Remedies
• Focused addressing
• Heavy load ? agents with free resources or agents
  soliciting contracts announce availability
• Light load ? agents with tasks or agents offering
  contracts announce availability.
• Audience restriction
• An agent negotiates with a subset of agents in
  the system.
• Ignoring outdated messages.

19
Decision Support Learning

• Creating an agent with a complete set of
  strategies (no learning) vs
• Acquisition of experience from the previous
  negotiations (learning).
• Genetic algorithms and genetic programming.
• Q-learning (reinforced learning).
• Other techniques.
• Learning is computationally expensive.

20
Decision Support Perfect Rationality vs Bounded
Rationality

• Two approaches
• Microeconomics ( e.g. Kreps1990, Varian1992
  and Raifa1982)
• and distributed artificial intelligence (DAI)
  (Rosenschein and Zlotkin1994, Durfee1994).
• Perfect rationality assumes that an agent can
  accurately model the environment and perform
  exact calculation for their decision-making.
• computation is complex and resource/time
  consuming.

21
Perfect Rationality vs Bounded Rationality
(contd)

• Bounded rationality means that resources are
  costly and bounded, and the model of the
  environment is not accurate
• e.g. the environment is dynamic and evolves.
• An agent has to
• decide how much computation to perform per
  task/contract
• choose a subset of tasks/ contracts available for
  consideration.

22
Combinatorial Aspects of Negotiation

• Negotiating with several agents or several
  marketplaces at a time.
• Computation complexity.

23
Automated Negotiations What's lacking?

• Why does current electronic commerce not widely
  support negotiations?
• Negotiation is difficult.
• Stumbling blocks.
• Need for a clear unambiguous ontology.
• Need for a strategy.

24
Ontologies

• Ontology is a way of categorizing objects such
  that they are semantically meaningful to a
  software agent.
• Must capture all important attributes of an
  object to allow for intelligent bargaining on
  both sides of a negotiation process
• Active research area in AI (see, for example,
  Sowa 1999).
• Existing tools
• KIF (Knoledge Intergande Format)
• Ontolingua
• DAML.

25
Strategy

• Game theory
• Treats negotiation from a mathematical
  prospective.
• Recommends a course of actions to a participant
  taking into account the opponents strategies and
  payoffs.
• Allows formulation of moves beyond pure price
  determination.

26
Strategy (contd)

• Disadvantages of the game-theoretic approach
• Assumes perfect information identically perceived
  by all bargaining parties
• Information is asymmetric, expectations are
  heterogeneous.
• Assumes perfect rationality of all players
• Computational agents operate under constraints,
  i.e. have bounded rationality.

27
Automated Negotiations What's lacking? (contd)

• Sophisticated strategies are mathematically
  complex and computationally expensive.
• Current software agents implement a fairly simple
  set of governing rules.
• Emerging complexity of the overall system (see,
  for example, Maes and Chavez 1994).
• Inadequate rule sets might lead to a disastrous
  system behavior.
• Program trading and 1987 stock market crush.

28
Agent Technology and Automated Negotiations

• Mobility to navigate among electronic
  marketplaces
• Intelligence to incorporate sophisticated
  numerical algorithms for portfolio optimization
  along with capability to learn and interpret the
  environment
• Agency to interact with other market entities
  (e.g. other agents).

29
Promising application areas

• Retail e-commerce
• Online auctions ?
• Electricity markets
• Bandwidth allocation
• Manufacturing planning scheduling in
  subcontracting networks
• Distributed vehicle routing among independent
  dispatch centers
• Electronic trading of financial instruments

30
Auction - Definitions

• An auction is a method of allocating scarce
  goods,
• a method that is based upon competition
• A seller wishes to obtain as much money as
  possible,
• A buyer wants to pay as little as necessary.
• An auction offers the advantage of simplicity in
  determining market-based prices.
• It is efficient in the sense that it usually
  ensures that
• resources accrue to those who value them most
  highly
• sellers receive the collective assessment of the
  value.
• The price is set the bidders.

31
When are Auctions Used?

• Auctions are useful when
• selling a commodity of undetermined quality
• the goods do not have a fixed or determined
  market value, in other words, when a seller is
  unsure of the price he can get.
• Choosing to sell an item by auctioning it off is
  
• more flexible than setting a fixed price
• less time-consuming and expensive than
  negotiating a price.
• Auctions can be used
• for single items such as a work of art
• and for multiple units of a homogeneous item such
  as gold or Treasury securities.

32
Prices

• The price is set not by the seller, but by the
  bidders.
• The seller sets the rules by choosing the type
  of auction to be used.
• The auctioneer doesn't often own the goods, but
  acts rather, as an agent for someone who does.
• The buyers frequently know more than the seller
  about the value of the item.
• A seller, not wanting to suggest a price first
  out of fear that his ignorance will prove costly,
  holds an auction to extract information he might
  not otherwise realize.

33
Bidder Valuations

• Reasons for bidding in the auction
• a bidder wishes to acquire goods for personal
  consumption (wine or art)
• a bidder wishes to acquire items for resale or
  commercial use.
• Private valuation
• Goods are acquired goods for personal
  consumption
• The bidder makes his own private valuation of the
  item for sale.
• All bidders have private valuations and tend to
  keep that information private.
• There would be little point in an auction if the
  seller knew already how much the highest
  valuation of an object will be.

34
Bidder Valuations (contd)

• Common valuation
• Goods are acquired goods for resale or commercial
  use
• An individual bid is predicated not only upon a
  private valuation reached independently, but also
  upon an estimate of future valuations of later
  buyers. Each bidder of this type tries (using the
  same measurements) to guess the ultimate price of
  the item.
• The item is really worth the same to all, but the
  exact amount is unknown
• Example
• Purchasing land for its mineral rights
• Each bidder has different information and a
  different valuation, but each must guess what
  price the land might ultimately bring.

35
Taxonomy of Auctions

• William Vickrey Vickrey established the basic
  taxonomy of auctions based upon the order in
  which prices are quoted and the manner in which
  bids are tendered. He established four major (one
  sided) auction types
• English Ascending-price, open-cry
• Dutch descending-price, open-cry,
• First-price, sealed bid,
• Vickrey or second-price, sealed bid.

36
English Auction

• The English auction
• the open-outcry auction or the ascending-price
  auction.
• "Here the auctioneer begins with the lowest
  acceptable price--the reserve price-- and
  proceeds to solicit successively higher bids from
  the customers until no one will increase the bid.
  The item is 'knocked down' (sold) to the highest
  bidder.
• Paul Milgrom

37
Dutch Auction

• In a Dutch auction, bidding starts at an
  extremely high price and is progressively lowered
  until a buyer claims an item.
• When multiple units are auctioned, normally more
  takers claim the item as price declines.
• The first winner takes his prize and pays his
  price
• Later winners pay less.
• When the goods are exhausted, the bidding is
  over.

38
First Price- Sealed Bid

• Sealed (not open-outcry like the English or Dutch
  varieties) and thus hidden from other bidders.
• A winning bidder pays exactly the amount he bid.
• Usually, (but not always) each participant is
  allowed one bid which means that bid preparation
  is especially important.
• a sealed-bid format has two distinct periods
• a bidding period in which participants submit
  their bids
• a resolution phase in which the bids are opened
  and the winner is determined (sometimes the
  winner is not announced).

39
Multiple Items in a Fist-Price, Sealed Bid Auction

• When multiple units are being auctioned, the
  auction is called "discriminatory" because not
  all winning bidders pay the same amount.
• In a first-price auction (one unit up for sale)
  each bidder submits one bid in ignorance of all
  other bids.
• The highest bidder wins and pays the amount he
  bid.
• In a "discriminatory auction, sealed bids are
  sorted from high to low, and items are awarded
  at highest bid price until the supply is
  exhausted.
• Winning bidders can (and usually do) pay
  different prices.

40
The Vickrey Auction

• The uniform second-price auction is commonly
  called the Vickrey auction.
• The bids are sealed, and each bidder is ignorant
  of other bids.
• The item is awarded to highest bidder at a price
  equal to the second-highest bid (or highest
  unsuccessful bid).
• winner pays less than the highest bid.
• Example
• Suppose bidder A bids 10, bidder B bids 15, and
  bidder C offers 20, bidder C would win, however
  he would only pay the price of the second-highest
  bid, namely 15.

41
The Vickrey Auction (contd)

• When auctioning multiple units, all winning
  bidders pay for the items at the same price
• the highest losing price.
• It seems obvious that a seller would make more
  money by using a first-price auction, but, in
  fact, that has been shown to be untrue.
• Bidders fully understand the rules and modify
  their bids as circumstances dictate.

42
Classification
Seller announces reserve price or some low
opening bid. Bidding increases progressively
until demand falls. Winning bidder pays highest
valuation. Bidder may re-assess evaluation during
auction.
English
Seller announces very high opening bid. Bid is
lowered progressively until demand rises to match
supply.
Dutch
43
Classification (contd)
Bids submitted in written form with no knowledge
of bids of others. Winner pays the exact amount
he bid.
First-price, sealed bid or discriminatory
Bids submitted in written form with no knowledge
of the bids of others. Winner pays the
second-highest amount bid.
Vickrey
44
Double Auction

• In this auction both sellers and buyers submit
  bids which are then ranked highest to lowest to
  generate demand and supply profiles.
• From the profiles, the maximum quantity
  exchanged can be determined by matching selling
  offers (starting with lowest price and moving up)
  with demand bids (starting with highest price and
  moving down).
• This format allows buyers to make offers and
  sellers to accept those offers at any particular
  moment.

45
More Auction Varieties

• In a sequential auction items are auctioned one
  at a time.
• A continuous double auction is one in which many
  individual transactions are carried on at a
  single moment and trading do not stop as each
  auction is concluded.
• Price formation mechanism for financial markets.
• In a parallel auction items are open for aution
  simultaneously and bidders may place their bids
  during a certain time period.
• In a combinatorial auction bidders place bids on
  combinations of items.
• Increase in overall auction complexity is
  stimulated by rapid development in the areas of
• Game theory
• Agent technology
• Electronic commerce (infrastructure).

46
Online Auctions

• Spawned by Research Communities
• AuctionBot (Michigan University)
• Kasbah (MIT Media Lab)
• e-Mediator (Tuomas Sandholm et al, Wasington
  University). ?
• Commercial
• eBay (www.ebay.com)
• OnSale (www.onsale.com) - defunct.

47
AuctionBot (University of Michigan)

• General purpose Internet auction server.
• Its users create new auctions by choosing from a
  selection of auction types and its parameters.
• User-specified
• Auction type (e.g. English, Dutch, etc.)
• Parameters
• Clearing times (reserve price, etc.)
• Method for resolving tie bids
• Number of sellers permitted.

48
AuctionBot (contd)

• Buyers sellers bid through multilateral
  distributive negotiation protocols.
• Advantage
• Provides an API for users to create own software
  agents to autonomously compete in AuctionBot
  marketplace.
• Users encode their own bidding strategies.

49
Kasbah (MIT MediaLab)

• Online multi-agent consumer-to-consumer
  transaction system.
• User that want to buy or sell an item proceed as
  follows
• Create an agent
• Give it some strategic directions
• Send off to a centralized marketplace.
• Proactive
• Agents seek out buyers / sellers and
• Negotiate on behalf of owners
• Obey users constraints.

50
Kasbah (contd)

• Goal
• Complete acceptable deal on behalf of a user
• Subject to set of user constraints
• Initial bidding (asking) price
• Lowest (highest) acceptable price
• Date to complete
• Restrictions on parties to negotiate with
• Price change over time.

51
Negotiation in Kasbah

• After buying agents and selling agents are
  matched
• Buying agents offer bid
• No restriction on time or price.
• Selling agents respond with binding yes or No.

52
Kasbah Negotiation Strategies
Buyer

• Anxious
• Linear
• Cool-headed
• Quadratic
• Frugal
• Exponential
• for increasing/decreasing its bid for a product
  over time.

Seller
53
Kasbah Trust Reputation (BBB)

• Better business bureau.
• Upon transaction completion both parties may rate
  other party, e.g.,
• Accuracy of product condition
• Completion of transaction.
• Agents use accumulated ratings
• Determine agents of owners who fall below
  specified reputation threshold gt no negotiation

54
eMediator Features

• http//ecommerce.cs.wustl.edu/emediator/
• E-commerce server.
• Driven by mobile agent technology.
• Performs a variety of e-commerce services.
• Has a configurable auction component that
  supports a variety of generalized combinatorial
  auctions, price setting mechanism, novel bid
  types.
• Has a leveled commitment contract optimizer that
  determines the optimal contract price and
  decommitting penalties.
• Has an exchange planner that enables unenforced
  anonymous exchanges by dividing the exchange into
  chunks and sequencing them to be delivered safely
  in alternation to buyer and seller.

55
eMediator Services

• eAuctionHouse
• Free-to-use third party auction site
• Wide range of customizable auction types
• Allows you to buy and sell items as well as to
  set up markets
• Allows bidding on combinations of items, and uses
  novel efficient algorithms for determining
  winners in this setting
• Allows bidding with price-quantity graphs
• Supports easy creation and safe on-site execution
  of mobile Java agents that can buy and sell goods
  on the server, and set up auctions
• Implemented in mostly in Java with some
  computationally expensive matching algorithms
  being written in C.

56
Optimal Winner Determination

• In a combinatorial auction individual bids by the
  same agent are joined by non-exclusive OR.
• Problem decide which bids win as to maximize the
  sum of the bid prices.
• Cannot be done in polynomial time (unless PNP).
• Approximate winner determination does not provide
  worst-case guarantee in polynomial time
• No polytime algorithm can guarantee an allocation
  within 1/(n1-e) bound from optimum for any egt0.
• Uses highly optimized tree-search-based
  heuristics
• Exponential worst case
• Scales up well in practice.
• Simplification possible through enforcing special
  bid structure
• XOR bids
• Stipulating maximal number of combinations to be
  excepted.

57
Price-Quantity Graphs

• Bidders can express continuous preferences, e.g.
  when a bidder buys a large quantity it will only
  take lower price per unit.
• Curves are piecewise linear for convenience of
  winner determination.
• In a single-sided auction with PQG bidding the
  winner determination proceeds as follows
• Sum the demand for every unit price
• Pick the aggregate solution that maximizes the
  unit price under the constraint that not more is
  demanded than is available
• Each bidder gets the amount that it bid at that
  unit price.
• Extensions to double auctions.

58
Mobile Agents

• User have agents participate in auctions on their
  behalf while their computers are offline.
• Agents execute on the agent dock which is on (or
  near) the host machine of the e-commerce server
• Gives mobile agents safe execution platform to
• Bid
• Set up auctions
• Travel to other auction sites
• Observe activity at various auctions.
• Reduces network latency
• Key issue in time-critical bidding.
• Uses commercial Mitsubishis Concordia agent
  dock http//www.meitca.com/HSL/Projects/Concordia

59
Concordia

• CONCORDIA is a framework for development and
  management of network-efficient mobile agent
  applications for accessing information on a
  device supporting Java.
• Concordia applications
• Process data at the data source
• Process data even if the user is disconnected
  from the network
• Access and deliver information across multiple
  networks (LANs, Intranets and Internet)
• Use wire-line or wireless communication
• Support multiple client devices, such as Desktop
  Computers, PDAs, Notebook Computers, and Smart
  Phones

60
HTML Interface

• Users instruct agents.
• Automatic generation of Java code for mobile
  agents before launching
• Allows non-programmers to create and launch their
  agents.

61
eMediator Services

• eCommitter, Leveled commitment contract optimizer
  
• Improves the efficiency of contracts by allowing
  decommitting
• Optimizes the contract price and the decommitment
  penalties for all contract parties based on a
  game-theoretic model
• Determines exactly when each party should
  decommit.

62
Contract Management

• Usually a contract is binding
• Cant undo old commitments to accommodate new
  events.
• Anticipated and true development
• An agent accrues information with time
• This may change his perception of the
  profitability of the contract
• E.g., tasks more costly than anticipated
• New offers more lucrative.
• Leveled commitment contracting protocol serves to
  alleviate the above
• Agents accommodate future events
• Option of unilateral decommit
• Decommitment penalty.

63
eMediator Services (contd)

• eExchangeHouse, a safe exchange planner
• Helps avoid non-delivery in exchanges
• Game-theoretic method for guaranteeing that each
  party is motivated to carry through with the
  exchange instead of vanishing
• Algorithms for chunking the exchange into parts
• Algorithms for sequencing the chunks to achieve
  safe exchange.
• Coalition formation support (coming soon)
• Electronic meeting place and discussion forum
• Efficient algorithms for coalition structure
  generation, and for dividing the coalition's
  payoffs

64
eMediator Services (contd)

• eVoter (coming soon)
• Nonmanipulable third party voting protocol
• Game-theoretically guarantees that voters are
  motivated to vote truthfully.
• Meta-auction (coming later)
• For finding out what is for sale on the Web.
• Reputation databases and algorithms (coming
  later)
• Collaborative rating of goods (coming later)

65
Future Directions

• Agents help buyer seller
• Combat info overload
• Expedite specific stages.
• First generation agents
• Create new markets
• Reduce transaction costs.

66
Future Directions (contd)

• Agent technologies need to better manage
• Ambiguous content
• Personalized preferences
• Complex goals
• Changing environments
• Disconnected parties.

67
Future Directions (contd)

• Standards
• Unambiguously universally define
• Goods services
• Consumer merchant profiles
• Value-added services
• Secure payment mechanisms
• Interbusiness electronic forms.

68
Future Directions (contd)

• Further
• New types of transactions
• Dynamic relationships among unknown parties.
• Create dynamic business partnerships that exist
  only as long as necessary.

69
Future Directions (contd)

• 3rd generation agent-mediated e-commerce.
• Markets with perfect efficiency.