The answers are in boldface.

1. Which one of the following is not an example of a possible application of multiagent systems?

   • Voice recognition
   • Automated manufacturing
   • Sensor fusion
   • Supply-chain automation
   • Robot coordination in a Mars mission

2. Which one of the following is the correct definition for agent? (in this class).

   • A computer system that is situated in some environment, and that is capable of autonomous action in this environment in order to meet its design objectives.
   • A reactive entity that inhabits an environment.
   • A system composed of discrete parts which we can refer to as beliefs, desires, and intentions.
   • A system capable of human-level intelligence.
   • A program.

3. A dynamic environment is one that

   • changes by itself
   • can be changed by agents
   • requires non-deterministic agents
   • supports mobile agents
   • is composed of dynamic agents

4. The formal definition of an agent includes the concept of a run. What is a run?

   • A sequence of environmental states and actions.
   • The set of all agents and their actions.
   • A list of the actions taken by an agent.
   • A list of the actions taken by all agents.
   • A function from actions to environmental states.

5. Two agents are said to be behaviorally equivalent if

   • They take the same action given the same history environmental states.
   • They are built with the same agent architecture.
   • They have the same set of inputs and the same set of possible actions.
   • They end up in the same final state at the same time.
   • They are both reactive.

6. Why are agents with state easier to implement than standard agents?

   • The states represent concepts that the programmer can relate to.
   • Agents with state are less expressive than standard agents.
   • Standard agents can only implement reactive behaviors.
   • Agents with state are FIPA-compliant.
   • They are not!

7. A deductive reasoning agent works by applying deduction rules to its knowledge base in an effort to

   • derive a statement that tells it to take some action.
   • prove that the action it has taken is the correct one.
   • find the relative benefit of taking each of its possible actions.
   • prove that it is better to just think rather than to take action.
   • create a long-term plan which will guide its future action choice.

8. In AGENT0, an agent has

   • capabilities, beliefs, commitments, and commitment rules.
   • beliefs, desires, and intentions.
   • a state to action mapping function.
   • desires, messages, and commitments.
   • plans, desires, joint intentions, and intention resolution rules.

9. You are building an robotic agent that will go into burning buildings and rescue people. Which one of the following might be one of the agent's intentions?

   • Reach the person at floor X by climbing the stairs in the East corner.
   • Save the people.
   • There is a fire on the third floor, and there is another robot on the second floor.
   • Fire can be extinguished using a fire extinguisher. There is one in each floor next to the elevator.
   • Take a step forward.

10. What is the difference between intending to do well on this test and desiring to do well on this test (assuming a BDI agent)?

    • the intention has specific actions associated with it.
    • there is no difference.
    • the desire can be held for a longer time.
    • the intention is stronger.
    • the desire is used for conflict resolution while the intention is used for filtering.

11. The subsumption architecture is composed of a set of behaviors and

    • an inhibition function.
    • a set of behavior rules.
    • a set of actions.
    • a set of desires.
    • nothing else.

12. The subsumption architecture was invented by

    • Rodney Brooks
    • John Searle
    • Marvin Minsky
    • Don Mills
    • Robert Wynford

13. How do you write the phrase "Cars have a license plate number." in RDF?

    • <rdf:Property ID="LicensePlateNumber">
        <rdfs:domain rdf:resource="http://jmvidal.cse.sc.edu/test#Cars"/>
        <rdfs:range rdf:resource="http://jmvidal.cse.sc.edu/test#LicenseNumber"/>
      </rdf:Property>
      

    • <rdf:Description ID="Cars">
        <rdf:Property ID="LicensePlate">
      </rdf:Description>
      

    • <rdf:Class ID="CarLicensePlates">
        <rdfs:Property about="#LicensePlate">
      </rdf:Class>
      

    • <Cars>
        <have>LicencePlate</have>
      </Cars>
      

    •  <rdf:Description ID="Cars">
          <rdf:type resource="http://jmvidal.cse.sc.edu/test#Class"/>
          <rdfs:subClassOf
              rdf:resource="http://jmvidal.cse.sc.edu/test#LicensePlates"/>
        </rdf:Description>
      

14. How do express the fact that there are Law students and Ethics students but there cannot be someone who is both a student of Law and Ethics, in DAML?

    • <daml:Class rdf:ID="LawStudent"/>
      <daml:Class rdf:ID="EthicsStudent">
        <daml:disjointWith rdf:resource="#LawStudent"/>
      </daml:Class>
      

    • <daml:Property rdf:ID="LawandEthics">
        <daml:domain rdf:resource="http://jmvidal.cse.sc.edu/test#Student"/>
        <daml:range rdf:resouce="rdfs:nil"/>
      </daml:Property>
      

    • <daml:Class rdf:ID="LawStudent"/>
        <rdfs:subClassOf rdf:resource="#Student"/>	  
      </daml:Class>
      <daml:Class rdf:ID="EthicsStudent"/>
        <rdfs:subClassOf rdf:resource="#Student"/>	  
      </daml:Class>
      

    • <daml:notEqual>
        <left>LawStudent</left>
        <right>EthicsStudent</right>
      </daml:notEqual>
      

    • <daml:Property rdf:ID="LawStudent"/>
      <daml:Property rdf:ID="EthicsStudent"/>
      <daml:notEqual>
        <daml:domain rdf:resource="#LawStudent"/>
        <daml:range rdf:resouce="#EthicsStudent"/>
      </daml:notEqual>
      

15. A DAML Restriction (daml:Restriction) is a

    • way to create an anonymous class that obeys certain constraints.
    • method for limiting users to certain semantic interpretations of the concepts presented in the ontology.
    • rule that limits the types of properties that a class can have.
    • way to map a class from one ontology to another one.
    • necessary but useless construct in order to make DAML backwards-compatible with RDF.

16. What is DAML used for?

    • Creating ontologies.
    • Describing agent behaviors.
    • Describing agent interactions.
    • Creating agent systems.
    • Deductive reasoning.

17. OWL is subdivided into three sublanguages of different level of expressiveness. This was done in order to make it easier for new users and

    • because of concerns about computational completeness.
    • to bridge the gap with DAML.
    • facilitate the integration with Web Services.
    • enable inter-operation with logical content languages.
    • to leverage the various levels of semantic meaning.

18. The main elements of a DAML-S service description are

    • ServiceProfile, ServiceModel, ServiceGrounding
    • ServiceLocation, ServiceInterphase
    • the WSDL description and the SOAP interface
    • ServiceActivation, ServiceInvocation, ServiceCanellation
    • Input, Outputs, Preconditions, Effects

19. DAML-S was created to be used in

    • describing web services at a semantic level.
    • describing ontologies about web services.
    • extending DAML to other languages.
    • describing the input/output types used in invoking a web service.
    • FIPA-compliant agent systems.

20. If you want to define a process to be a DAML-S atomic process, you would do this by writing DAML-S code that says that the process

    • is a subclass of damls:AtomicProcess
    • has the property damls:AtomicProcess
    • has the property damls:composedOf that points to rdf:nil
    • is a type of damls:CompositeProcess
    • is a type of process that has only one damls:Action

21. A DAML-S CompositeProcess is roughly equivalent to

    • a small programming language.
    • the DAML:class
    • the rdf:subPropertyOf
    • a set of atomic processes to be executed in some order.
    • the RDF:LargeUnit

22. In netlogo, the without-interruption primitive is used to

    • force agents to only run one at a time.
    • prevent the user from stopping the agents while they are running.
    • run long processes.
    • prevent the user from changing the code on a running system.
    • give one agent higher priority than the other ones.

23. In netlogo, a function that returns a value is called a

    • to-report
    • to-return
    • to (but with a return statement)
    • eval-fun
    • with-values

24. In netlogo, how would I ask all the red agents to face east?

    • ask turtles with [color = red] [set heading 90]
    • ask red-color turtles [face east]
    • ask turtles with [color = red] [set-facing east]
    • ask turtles-with [color = red] [face east]
    • ask turtles [if (color = red) then set heading 90]

25. Which one of the following is an example of a performative?

    • ask
    • drop
    • perform
    • imagine
    • forward

26. The Knowledge Interchange Format is roughly:

    • first-order logic in Lisp format.
    • RDF with some first-order relations.
    • a subset of Lisp.
    • a way to re-write DAML.
    • KQML without the performatives.

27. The FIPA-ACL protocol element denotes

    • the interaction protocol being used.
    • the encoding protocol being used.
    • the knowledge format protocol being used.
    • the level of urgency.
    • the type of agent initiating the conversation.

28. The FIPA-ACL inform-if performative is used to

    • ask another agent if a statement is true or false.
    • ask another agent to tell us something, but only if a particular clause evaluates true.
    • ask one agent to tell something to a third agent.
    • ask another agent to keep checking the value of a particular clause.
    • tell another agent if something is true.

29. The main reason we need a language such as FIPA-ACL is that

    • it tells an agent what to do with the knowledge it receives.
    • it encodes all types of information.
    • it is pro-active.
    • it enables the use of DAML-based multiagent systems.
    • it adds a semantic layer to the OSI networking model.

30. The semantics of FIPA-ACL performatives are defined using a formal language which uses

    • beliefs and desires
    • reactive methodology
    • modulus arithmetical propositions
    • commitments, intentions, and post-conditions
    • calculus

31. In a FIPA architecture the AMS is responsible for

    • white pages service
    • yellow pages service
    • message passing
    • message routing across platforms
    • message termination

32. AUML is like UML's sequence diagram except that AUML

    • allows the representation of multiple (possible) time-lines.
    • the messages must be FIPA-ACL performatives.
    • there is no goal state(s).
    • supports the representation of messages from one agent/object to another one.
    • uses beliefs and desires to represent the intended semantics of the message.

33. What are the Pareto optimal strategies of the game:

    	A
    B		m	n
    	o	3,4	0,9
    	p	2,7	6,1

    • (o,m), (o,n), (p,n)
    • (o,m), (o,n), (p,n), (p,m)
    • (o,m), (o,n)
    • (o,m)
    • (p,n)

34. What are the pure Nash equilibrium strategies of the game:

    	A
    B		m	n
    	o	3,4	0,9
    	p	2,7	6,1

    • There aren't any.
    • (o,m), (o,n), (p,n)
    • (o,m), (o,n)
    • (o,m)
    • (p,n)

35. What are the pure Nash equilibrium strategies of the game:

    	A
    B		m	n
    	o	3,4	0,3
    	p	2,7	9,1

    • (o,m)
    • (o,n)
    • (o,n)
    • (p,m)
    • (p,n)

36. Why is the fact that most games have multiple equilibria important to us as multiagent systems designers?

    • We must develop coordination techniques so agents can agree on which one to play.
    • It means that multiagent systems are more likely to be developed since agents are more likely to agree on a strategy.
    • They are unimportant since all we care about is what the agents actually play.
    • Each equilibrium represents one possible multiagent protocol implementation.
    • The decisions of many agents are proportional to the number of equilibria.

37. What is the social welfare strategy of the following game matrix

    	A
    B		m	n	o	p
    	m	1,1	3,2	1,3	2,4
    	n	5,1	3,1	4,1	5,1
    	o	4,2	1,1	4,4	2,5
    	p	4,1	2,2	4,3	2,4

    • (o,o)
    • (n,m)
    • (n,p)
    • (p,p)
    • (o,m)

38. What is the dominant strategy in the Prisoner's dilemma?

    • Defect
    • Cooperate
    • Tit-for-Tat
    • There isn't one.
    • Defect, then Tit-for-Tat

39. What did Axelrod conclude about playing Tit-for-Tat in the iterated Prisoner's Dilemma game?

    • That it is the most successful strategy when many players play each other on many games.
    • That it is the best strategy to play against any opponent.
    • That there is a good probability that the other player will be playing Tit-for-Tat.
    • That, as far as we know, there is no way to beat it on one-on-one play against it.
    • That it is a good strategy to play as long as all your opponents are willing to cooperate.

40. A social convention

    • limits the available strategies.
    • limits the available actions.
    • imposes limits on deliberation time.
    • imposes limits on the possible histories.
    • eliminates the need for cooperation.