Adaptive interaction using an adaptive agent-oriented software architecture

• US 6,594,684 B1
• Filed: 07/24/2000
• Issued: 07/15/2003
• Est. Priority Date: 06/15/1998
• Status: Expired due to Term

• Current Assignee: iAnywhere Solutions Incorporated (SAP SE)
• Original Assignee: Dejima, Inc. (SAP SE)
• Inventors: Amamiya, Makoto, Hodjat, Babak, Savoie, Christopher J.
• Primary Examiner(s): Coulter, Kenneth R.
• Application Number:
  US09/624,763
• Time in Patent Office: 1,086 Days
• Field of Search: 709/202, 706/10
• US Class Current: 709/202
• CPC Class Codes: G06N 5/043 Distributed expert systems; Blackboards

Abstract

Method and agent network architecture for processing a subject message, where each agent has a view of its own domain of responsibility. An initiator agent which receives a user-input request and does not itself have a relevant interpretation policy, queries its downchain agents whether the queried agent considers such message to be in its domain of responsibility. Each queried agent recursively determines whether it has an interpretation policy of its own that applies to the request, and if not, further queries its own further downchain neighboring agents. The further agents eventually respond to such further queries, thereby allowing the first-queried agents to respond to the initiator agent. The recursive invocation of this procedure ultimately determines one or more paths through the network from the initiator agent to one more more leaf agents. The request is then transmitted down the path(s), with each agent along the way taking any local action thereon and passing the message on to the next agent in the path. In the event of a contradiction, the network is of ten able to resolve many of such contradictions according to predetermined algorithms. If it cannot resolve a contradiction automatically, it learns new interpretation policies necessary to interpret the subject message properly. Such learning preferably includes interaction with the user (but only to the extent necessary), and preferably localizes the learning close to the correct leaf agent in the network.

Claims

1. A computer-implemented method for use with a subject message, for use further with a network of agents each having a view of its own domain of responsibility, comprising the steps of a first one of said agents:

receiving a command to accept said subject message; and

in response to said step of receiving a command, learning that messages having characteristics of said subject message are within the domain of responsibility of said first agent,

further comprising the step of producing an output of said network in response to said subject message.

2. A method according to claim 1, further comprising the step of , in response to said step of receiving a command, said first agent instructing a second group of at least one agent downchain of said first agent to handle said subject message.

3. A method according to claim 2, wherein said step of instructing a second group of at least one agent to handle said subject message comprises the step of instructing at least one agent in said second group to handle said subject message.

4. A method according to claim 2, further comprising the step of , in response to said step of receiving a command, said first agent resolving among at least two groups of agents downchain of said first agent to identify said second group of at least one agent.

5. A method according to claim 4, wherein said step of resolving comprises the step of evaluating how recently said first agent instructed an agent in each of said downchain groups to handle a prior message.

6. A method according to claim 4, wherein said step of resolving comprises the steps of:

comparing quality levels of responses previously received from said groups of downchain agent; and

evaluating how recently said first agent instructed an agent in each of said downchain groups to handle a prior message.

7. A method according to claim 6, wherein said step of resolving further comprises the step of causing a user to be asked to select among said groups of downchain agents.

8. A method according to claim 4, further comprising the step of said first agent learning that future messages having a characteristic of said subject message are within the domain of responsibility of said second group of at least one agent.

9. A method according to claim 4, further comprising the steps of , in response to said a second agent in said second group of agents receiving said instruction from said first agent, said second agent learning that messages having characteristics of said subject message are within the domain of responsibility of said second agent.

10. A method according to claim 9, wherein said step of producing an output of said network comprises the step of said second agent taking local steps to produce an output of said network in response to said step of said second agent receiving a command.

11. A method according to claim 4, wherein said step of resolving comprises the step of causing a user to be asked to select among at least two groups of agents downchain of said first agent.

12. A method according to claim 1, wherein said step of producing an output of said network comprises the step of said first agent taking local steps to produce an output of said network in response to said step of receiving a command.

13. A method according to claim 1, wherein said command is received from an agent upchain of said first agent, further comprising the step of , in response to said step of receiving a command, sending a command to said upchain agent to unlearn that messages having characteristics of said subject message are within the domain of responsibility of said first agent.

14. A method according to claim 13, wherein prior to said step of sending a command to said upchain agent to unlearn, said upchain agent had an interpretation policy that messages having characteristics of said subject message are within the domain of responsibility of said first agent, further comprising the step of said upchain agent, in response to receipt of said command to said upchain agent to unlearn, unlearning that messages having characteristics of said subject message are within the domain of responsibility of said first agent.

15. A method according to claim 13, wherein prior to said step of sending a command to said upchain agent to unlearn, said upchain agent did not have an interpretation policy that messages having characteristics of said subject message are within the domain of responsibility of said first agent, further comprising the step of said upchain agent ignoring said command to said upchain agent to unlearn that messages having characteristics of said subject message are within the domain of responsibility of said first agent.

16. A method according to claim 1, wherein said output is human perceptible.

17. A method according to claim 1, wherein said subject message includes user input.

18. A computer-implemented method for use with a network of agents, comprising the step of a first one of said agents sending an instruction to at least one recipient agent in said network instructing each of said recipient agent(s) to unlearn any interpretation policy that it has which calls for messages having a predetermined characteristic to be sent to said first agent.

19. A method according to claim 18, wherein said step of sending a message to at least one recipient agent comprises the step of sending said message to all agents which are upchain of said first agent in said network.

20. A method according to claim 19, wherein said step of sending a message to at least one recipient agent comprises the step of sending said message to all agents in said network which are known to said first agent.

21. A computer-implemented method for use with a network of agents, comprising the steps of a first agent in said network:

receiving from another agent a first interpretation problem for evaluating whether a subject message is within a domain of responsibility of said first agent;

storing said first problem accessibly to said first agent; and

subsequently

receiving an instruction calling on said first agent to forget said first problem.

22. A method according to claim 21, further comprising the step of said first agent deleting said first problem in response to said step of receiving.

23. A method according to claim 21, wherein each agent in said network of agents has a view of its own domain of responsibility, further comprising the steps of said first agent:

making a determination of whether said subject message is within said first agent'"'"'s domain of responsibility; and

returning said determination to said other agent.

24. A method according to claim 23, wherein said step of making a determination comprises the step of making a determination of whether said subject message is within said first agent'"'"'s local domain of responsibility.

25. A method according to claim 23, wherein said step of making a determination comprises the step of making a determination of whether said subject message is within the domain of responsibility of any agents downchain of said first agent.

26. A method according to claim 23, wherein said step of making a determination of whether said subject message is within the domain of responsibility of any agents downchain of said first agent, comprises the step of querying at least one agent downchain of said first agent whether the queried agent considers said subject message to be in the queried agent'"'"'s domain of responsibility.

27. A method according to claim 23, wherein said step of receiving an instruction calling on said first agent to forget said first problem occurs after said step of returning said determination to said other agent.

28. A method according to claim 23, further comprising the step of said first agent, after said step of returning said determination, receiving a command to process said subject message.

29. A method according to claim 28, wherein said instruction calling on said first agent to forget said first problem comprises said command to process said subject message.

30. A method according to claim 28, further comprising the step of said first agent taking local steps to produce an output of said network in response to said step of receiving a command to process said subject message.

31. A method according to claim 21, wherein said instruction calling on said first agent to forget said first problem comprises a FORGET-PROBLEM performative.

32. A method according to claim 21, further comprising the step of said first agent, in response to said step of receiving an instruction, sending an instruction to at least one further agent in said network instructing each of said further agent(s) to forget said first problem.

33. A method according to claim 32, wherein said step of sending an instruction to at least one further agent comprises the step of said first agent sending said message to all agents which have received said first problem from said first agent.

34. A method according to claim 32, wherein said step of sending an instruction to at least one further agent comprises the step of said first agent sending said message to all agents in said network which are downchain of said first agent.

35. A method according to claim 21, wherein said subject message includes user input.

36. A computer-implemented method for use with a network of agents, comprising the steps of:

propagating a plurality of interpretation problems through said network for identifying at least one path that each such problem should travel through said network, each agent that receives one of said problems performing steps which include storing the problem; and

sending an instruction to at least one recipient agent in said network instructing each of said recipient agent(s) to forget said one problem.

37. A method according to claim 36, wherein said step of sending a message to at least one recipient agent comprises the step of sending said message to all agents which have received said one problem.

38. A method according to claim 37, wherein said step of sending a message to at least one recipient agent comprises the step of sending said message to all agents in said network.

39. Computing apparatus comprising a network of a plurality of agents, each of said agents having a view of its domain of responsibility, and each given one of said agents including:

an interpreter which determines whether a subject message is within the given agent'"'"'s domain of responsibility; and

learning means which adapts the given agent'"'"'s view of its domain of responsibility in response to contradiction resolution.

40. Apparatus according to claim 39, wherein said apparatus further comprises additional agents not in said network.

41. Apparatus according to claim 40, wherein said network and said additional agents collectively define a further network.

42. Apparatus according to claim 39, wherein the given agent'"'"'s domain of responsibility includes a local domain of responsibility of the given agent.

43. Apparatus according to claim 39, wherein the given agent has at least one neighboring agent downchain of said given agent, and wherein the given agent'"'"'s domain of responsibility includes the domains of responsibility of all of said downchain agent(s).

44. Apparatus according to claim 43, wherein the given agent'"'"'s domain of responsibility further includes a local domain of responsibility of the given agent.

45. Apparatus according to claim 43, wherein the interpreter in said given agent includes means for querying at least one agent downchain of said given agent whether the queried agent considers said subject message to be in the queried agent'"'"'s domain of responsibility.

46. Apparatus according to claim 45, wherein the interpreter in said given agent includes means for resolving contradictory responses received from the queried agents.

47. Apparatus according to claim 46, wherein said learning means is responsive to said resolving means.

48. Apparatus according to claim 46, wherein said resolving means includes means for discarding query responses which do not satisfy threshold quality criteria.

49. Apparatus according to claim 48, wherein query responses each have associated therewith a respective indication of a priority level of the agent which originated the query response, and wherein said threshold quality criteria require a priority level above a predetermined threshold priority level.

50. A method according to claim 46, wherein said resolving means includes means for evaluating how recently said given agent instructed each of said downchain agents to handle a prior message.

51. Apparatus according to claim 46, wherein said resolving means includes:

means for comparing quality levels of responses previously received from said downchain agent(s); and

means includes means for evaluating how recently said given agent instructed each of said downchain agents to handle a prior message.

52. Apparatus according to claim 46, wherein said resolving means includes means for causing a user to be asked to select among at least two agents downchain of said given agent.

53. Apparatus according to claim 43, wherein one of said agents further includes:

storage means containing identifications of the downchain neighboring agents of said given agent; and

means for adapting said identifications in response to messages received from other agents.

54. Apparatus according to claim 39, wherein said learning means adapts the given agent'"'"'s view of its domain of responsibility also in response to feedback regarding prior interpretations made by said interpreter.

55. Apparatus according to claim 54, wherein said feedback originates from a user.

56. Apparatus according to claim 39, wherein a first one of said agents further includes a specialized processing unit which, in response to receipt of a predetermined performative, performs local steps toward producing an output of said network.

57. Apparatus according to claim 56, wherein a second one of said agents further includes a specialized processing unit which, in response to receipt of said predetermined performative, performs local steps toward producing an output of said network,

the local steps performed by the specialized processing unit in response to receipt of said predetermined performative being different from the local steps performed by said first processing unit in response to receipt of said predetermined performative.

58. Apparatus according to claim 39, wherein said computing apparatus comprises a single processor, all of said agents executing on said single processor.

59. Apparatus according to claim 39, wherein said computing apparatus comprises a plurality of processors, said agents being spread over said plurality of processors.

60. Apparatus according to claim 39, wherein said computing apparatus comprises a plurality of computers communicating over the internet, said agents being spread over said plurality of computers.

61. Apparatus according to claim 39, wherein one of said agents is implemented in hardware.

62. A computer-implemented method for use with a subject message, for use further with a network of agents each having a view of its own domain of responsibility, comprising the steps of a first one of said agents:

receiving from an upchain agent a query inquiring whether said subject message is within the domain of responsibility of said first agent;

determining whether said subject message is within said first agent'"'"'s domain of responsibility;

and where said subject message is within said first agent'"'"'s domain of responsibility, returning a response to said upchain agent claiming said subject message.

63. A method according to claim 62, wherein said step of determining comprises the step of determining whether said subject message is within said first agent'"'"'s local domain of responsibility.

64. A method according to claim 62, wherein said step of determining comprises the step of determining whether said subject message is within the domain of responsibility of any agents downchain of said first agent.

65. A method according to claim 64, wherein said step of determining whether said subject message is within the domain of responsibility of any agents downchain of said first agent, comprises the step of querying at least one agent downchain of said first agent whether the queried agent considers said subject message to be in the queried agent'"'"'s domain of responsibility.

66. A method according to claim 65, wherein said step of determining whether said subject message is within the domain of responsibility of any agents downchain of said first agent, further comprises the step of said first agent receiving, in response to said step of querying, only non-contradictory claim(s) of message ownership from at least one of said queried agent(s).

67. A method according to claim 66, wherein said step of determining whether said subject message is within the domain of responsibility of any agents downchain of said first agent, further comprises the step of said first agent:

receiving at least one query response from said queried agent(s); and

evaluating for contradiction only those of said query responses which satisfy threshold quality criteria.

68. A method according to claim 67, wherein each of said query responses has associated therewith an indication of a priority level of the agent which originated said query response, and wherein said threshold quality criteria require a priority level above a predetermined threshold priority level.

69. A method according to claim 68, wherein each of said query responses further has associated therewith an indication of confidence with which the agent originating the query response claims said subject message.

70. A method according to claim 62, further comprising the steps of said first agent, after said step of returning a response:

receiving a command to process said message;

producing an output of said network in response to said step of receiving a command.

71. A method according to claim 62, further comprising the steps of said first agent, after said step of returning a response:

receiving a command to process said message; and

forwarding to at least one further agent downchain of said first agent, a further command to process said message.

72. A method according to claim 71, further comprising the step of said first agent taking local steps to produce an output of said network in response to said step of receiving a command.

73. A method according to claim 62, wherein said network of agents is organized as a hyperstructure of agents.

74. A method according to claim 62, wherein said subject message includes user input.

75. A method according to claim 74, further comprising the steps of said first agent, after said step of returning a response:

receiving a command to process said message; and

taking steps to produce a user-perceptible output of said network in response to said step of receiving a command.

76. A method according to claim 74, further comprising the steps of said first agent, after said step of returning a response:

receiving a command to process said message;

forwarding to at least one further agent downchain of said first agent, a further command to process said message; and

taking local steps to produce a user-perceptible output of said network in response to said step of receiving a command.

77. A computer-implemented method for use with a subject message, for use further with a network of agents each having a view of its own domain of responsibility, comprising the steps of a first one of said agents:

receiving from an upchain agent a query inquiring whether said subject message is within the domain of responsibility of said first agent;

determining whether said subject message is within said first agent'"'"'s domain of responsibility;

and where said subject message is not within said first agent'"'"'s local domain of responsibility and said first agent has no downchain agents, returning a response to said upchain agent disclaiming said subject message.

78. A method according to claim 77, further comprising the step of a further agent in said network of agents claiming said subject message.

79. A method according to claim 77, further comprising the steps of said first agent, after said step of returning a response:

receiving a command to accept said subject message;

producing an output of said network in response to said step of receiving a command.

80. A method according to claim 77, further comprising the steps of said first agent, after said step of returning a response:

receiving a command to accept said subject message; and

learning that messages having characteristics of said subject message are within the domain of responsibility of said first agent.

81. A method according to claim 80, further comprising the step of said first agent taking local steps to produce an output of said network in response to said step of receiving a command.

82. A method according to claim 77, further comprising the step of producing an output of said network in response to said step of receiving a query.

83. A method according to claim 82, wherein said output is human perceptible.

84. A method according to claim 77, wherein said subject message includes user input.

85. A computer-implemented method for use with a subject message, for use further with a network of agents each having a view of its own domain of responsibility, comprising the steps of a first one of said agents:

receiving from an upchain agent a query inquiring whether said subject message is within the domain of responsibility of said first agent;

and where said subject message is not within said first agent'"'"'s local domain of responsibility but said first agent has further agents downchain of said first agent, querying at least one of said downchain agents whether the downchain agent considers said subject message to be in the downchain agent'"'"'s domain of responsibility.

86. A method according to claim 85, further comprising the steps of:

said first agent receiving, in response to said step of querying, only non-contradictory claim(s) of message ownership from at least one of said queried agent(s); and

in response thereto,

returning a response to said upchain agent claiming said subject message.

87. A method according to claim 86, further comprising the steps of said first agent, after said step of returning a response:

receiving a command to process said subject message;

producing an output of said network in response to said step of receiving a command.

88. A method according to claim 86, wherein said step of first said agent receiving only non-contradictory claim(s) of message ownership comprises the step of said first agent receiving first and second claims of ownership of said subject message in response to said step of querying, said first claim of ownership identifying a first focus of said subject message and said second claim of ownership identifying a second focus of said subject message different from said first focus of said subject message.

89. A method according to claim 85, further comprising the steps of said first agent:

receiving at least one query response from said queried agent(s); and

discarding all of said query responses which fail threshold quality criteria;

said step of discarding leaving only non-contradictory claim(s) of message ownership from at least one of said queried agent(s); and

in response thereto,

returning a response to said upchain agent claiming said subject message.

90. A method according to claim 85, further comprising the steps of:

said first agent receiving, in response to said step of querying, at least one response indicating that one of said downchain agents has a contradiction, and in response thereto,

said first agent returning a response to said upchain agent indicating that said first agent has a contradiction.

91. A method according to claim 90, further comprising the steps of said first agent, after said step of returning a response:

receiving a command to process said message; and

in response thereto,

forwarding to at least said one downchain agent, a further command to process said message.

92. A method according to claim 91, further comprising the step of said one downchain agent resolving its contradiction in response to receipt of said further command to process said message.

93. A method according to claim 92, further comprising the step of said first agent and said one downchain agent each taking respective local steps to produce an output of said network in response to said step of receiving a command.

94. A method according to claim 91, further comprising the step of said first agent taking local steps to produce an output of said network in response to said step of receiving a command.

95. A method according to claim 85, further comprising the steps of:

said first agent receiving, in response to said step of querying, at least two responses claiming contradictory ownership of said subject message, and in response thereto,

said first agent returning a response to said upchain agent indicating that said first agent has a contradiction.

96. A method according to claim 95, further comprising the steps of said first agent, after said step of returning a response:

receiving a command to process said message; and

resolving said contradiction in response to said step of receiving a command.

97. A method according to claim 96, wherein said step of resolving said contradiction comprises the step of identifying a second agent downchain of said first agent as being eligible to receive said subject message,

further comprising the step of said first agent forwarding to said second agent, further command to process said message.

98. A method according to claim 97, further comprising the step of each of said first and second agents taking respective local steps to produce an output of said network in response to said step of receiving a command.

99. A method according to claim 96, further comprising the step of said first agent taking local steps to produce an output of said network in response to said step of receiving a command.

100. A method according to claim 85, further comprising the steps of said first agent:

receiving at least one response in response to said step of querying; and

where none of said responses claim said subject message, returning a response to said upchain agent disclaiming said subject message as being within said first agent'"'"'s domain of responsibility.

101. A method according to claim 85, further comprising the steps of said first agent:

receiving at least one response in response to said step of querying;

evaluating said responses to determine whether any claim said subject message and also satisfy predetermined quality criteria; and

where none of said responses claim said subject message and also satisfy said predetermined quality criteria, returning a response to said upchain agent disclaiming said subject message as being within said first agent'"'"'s domain of responsibility.

102. A method according to claim 85, further comprising the steps of said first agent:

receiving at least one response in response to said step of querying; and

returning a response to said upchain agent disclaiming said subject message as being within said first agent'"'"'s domain of responsibility; and

after said step of returning a response,

receiving a command to accept said subject message;

producing an output of said network in response to said step of receiving a command.

103. A method according to claim 85, further comprising the steps of said first agent:

receiving at least one response in response to said step of querying; and

returning a response to said upchain agent disclaiming said subject message as being within said first agent'"'"'s domain of responsibility; and

after said step of returning a response,

receiving a command to accept said subject message; and

in response to said step of receiving a command, learning that messages having characteristics of said subject message are within the domain of responsibility of said first agent.

104. A method according to claim 103, further comprising the steps of, in response to said step of receiving a command, said first agent:

identifying a second agent downchain of said first agent as being eligible to receive said subject message; and

forwarding to said second agent a further command to process said subject message.

105. A method according to claim 103, further comprising the step of said first agent taking local steps to produce an output of said network in response to said step of receiving a command.

106. A method according to claim 104, further comprising the steps of, in response to said second agent receiving said further command to process said subject message, said second agent learning that messages having characteristics of said subject message are within the domain of responsibility of said second agent.

107. A method according to claim 104, wherein said step of identifying a second agent comprises the step of causing a user to be asked to select among at least two agents downchain of said first agent.

108. A method according to claim 106, further comprising the steps of, in response to said second agent receiving said further command to process said subject message, said second agent sending a command to said first agent to unlearn that messages having characteristics of said subject message are within the domain of responsibility of said first agent.

109. A method according to claim 106, further comprising the step of said second agent taking local steps to produce an output of said network in response to said step of said second agent receiving a command.

110. A method according to claim 85, further comprising the step of a further agent in said network of agents claiming said subject message.

111. A method according to claim 85, further comprising the step of producing an output of said network in response to said subject message.

112. A method according to claim 111, wherein said output is human perceptible.

113. A method according to claim 85, further comprising the steps of:

said first agent receiving, in response to said step of querying, at least two responses claiming contradictory ownership of said subject message, and in response thereto,

said first agent returning a response to said upchain agent indicating that said first agent has a contradiction.

114. A method according to claim 85, wherein said subject message includes user input.

115. A computer-implemented method for use with a network of agents including at least one agent having a plurality of downchain neighboring agents, comprising the steps of:

initiating a command into said network;

each agent receiving said command, which agent has exactly one downchain neighboring agent, forwarding said command to said exactly one downchain neighboring agent; and

each agent receiving said command, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said command and forwarding said command to each agent in said subset,

said network producing an output of said network in response to said step of initiating.

116. A method according to claim 115, wherein said step of determining a subset of said plurality comprises the step of determining a proper subset of said plurality.

117. A method according to claim 115, further comprising the step of at least one agent receiving said command taking local steps to produce an output of said network in response to said command.

118. A method according to claim 115, wherein said steps of initiating and forwarding collectively identify a first path taken by said command through said network,

and wherein at least a first agent in said first path takes local steps to produce an output of said network in response to said command.

119. A method according to claim 118, wherein said steps of initiating and forwarding further collectively identify a second path taken by said command through said network,

and wherein at least a second agent in said second path takes local steps to produce an output of said network in response to said command.

120. A method according to claim 119, wherein said first agent is not in said second path and said second agent is not in said first path.

121. A method according to claim 119, wherein at least one of said outputs of said network are human perceptible.

122. A method according to 115, further comprising the step of a first agent which receives said command learning that messages having characteristics of a subject message identified with said command are within the domain of responsibility of said first agent.

123. A method according to claim 122, further comprising the step of said first agent notifying an upchain neighboring agent in said network to unlearn any interpretation policy that messages having said characteristics are within the domain of responsibility of said first agent.

124. A method according to claim 122, comprising the step of each given agent in said network which receives said command learning that messages having said characteristics are within the domain of responsibility of the given agent.

125. A method according to claim 124, further comprising the step of each given agent which receives said command notifying other agents to unlearn any interpretation policy that messages having characteristics of a subject message identified with said command are within the domain of responsibility of said given agent.

126. A method according to claim 115, further comprising the step of a first agent which receives said command, which first agent has a plurality of downchain neighboring agents, learning that messages having characteristics of a subject message identified with said command are within the domain of responsibility of each of the agents in the subset of agents determined in said step of determining.

127. A method according to claim 115, wherein said step of each agent receiving said command, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said command, comprises the step of:

each given agent receiving said command, which given agent has a plurality of downchain neighboring agents, consulting a prior determination, previously stored accessibly to the given agent, of the subset of said plurality which should receive said command from the given agent.

128. A method according to claim 115, wherein said step of each agent receiving said command, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said command, comprises the step of:

one of such agents evaluating how recently it instructed an agent in said plurality of downchain neighboring agents to handle a prior message.

129. A method according to claim 115, wherein said step of each agent receiving said command, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said command, comprises the step of one of such agents:

comparing quality levels of responses previously received from agents in said plurality of downchain neighboring agents relative to a message associated with said command.

130. A method according to claim 115, wherein said step of each agent receiving said command, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said command, comprises the step of one of such agents:

comparing quality levels of responses previously received from agents in said plurality of downchain neighboring agents relative to a message associated with said command; and

evaluating how recently it instructed an agent in said plurality of downchain neighboring agents to handle a prior message.

131. A method according to claim 115, wherein said output of said network is human perceptible.

132. A method according to claim 115, wherein said step of initiating comprises the step of receiving an indication from a user that a prior response of said network to a prior problem was unsatisfactory.

133. A method according to claim 132, wherein said step of producing an output comprises the step of producing a human perceptible output.

134. A method according to claim 132, wherein said step of determining a subset comprises the step of causing a user to be asked to select said subset from among said plurality of downchain neighboring agents.

135. A method according to claim 115, wherein said step of each agent receiving said command, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said command, includes the step of:

one of such agents causing a user to be asked to select among said plurality of downchain neighboring agents.

136. A method according to claim 135, wherein said step of each agent receiving said command, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said command, further comprises the step of:

one of such agents evaluating how recently it instructed an agent in said plurality of downchain neighboring agents to handle a prior message.

137. A method according to claim 135, wherein said step of each agent receiving said command, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said command, further comprises the step of one of such agents:

comparing quality levels of responses previously received from agents in said plurality of downchain neighboring agents relative to a message associated with said command.

138. A computer-implemented method for use with a network of agents, each agent in said network having a view of its own domain of responsibility, comprising the steps of:

initiating a subject message into said network a first time;

each agent receiving said subject message the first time from a respective upchain neighboring agent the first time, evaluating whether said subject message is within the agent'"'"'s domain of responsibility and returning an indication of such evaluation to said upchain neighboring agent;

re-initiating said subject message into said network for a second time; and

each agent receiving said subject message the second time resolving any contradictions it has about whether said subject message is within the agent'"'"'s domain of responsibility.

139. A method according to claim 138, wherein said step of each agent receiving said subject message the first time evaluating whether said subject message is within the agent'"'"'s domain of responsibility, comprises the step of:

a first one of said agents receiving said subject message the first time making a determination of whether said subject message is within the local domain of responsibility of said first agent.

140. A method according to claim 138, wherein a first one of said agents receiving said subject message the first time has at least one downchain neighboring agent, and wherein said step of each agent receiving said subject message the first time evaluating whether said subject message is within the agent'"'"'s domain of responsibility, comprises the step of:

said first agent making a determination of whether said subject message is within the domain of responsibility of any of the downchain neighboring agents of said first agent.

141. A method according to claim 140, wherein said step of said first agent making a determination of whether said subject message is within the domain of responsibility of any of the downchain neighboring agents of said first agent comprises the step of said first agent consulting a database of interpretation policies local to said first agent.

142. A method according to claim 140, wherein said step of said first agent making a determination of whether said subject message is within the domain of responsibility of any of the downchain neighboring agents of said first agent comprises the step of said first agent querying at least one of the downchain neighboring agents of said first agent whether the queried agent considers said subject message to be in the queried agent'"'"'s domain of responsibility.

143. A method according to claim 138, wherein said step of evaluating whether said subject message is within the agent'"'"'s domain of responsibility comprises the steps of said agent classifying said subject message into one of a plurality of categories including:

(a) the agent considers the subject message to be within the agent'"'"'s domain of responsibility,

(b) the agent considers the subject message to be not within the agent'"'"'s domain of responsibility, and

(c) the agent has a contradiction about whether the subject message is within the agent'"'"'s domain of responsibility.

144. A method according to claim 143, wherein a first one of said agents receiving said subject message the first time has at least one downchain neighboring agent, and wherein said step of each agent receiving said subject message the first time evaluating whether said subject message is within the agent'"'"'s domain of responsibility, further comprises the steps of said first agent:

receiving at least one query response from said at least one downchain neighboring agent; and

discarding as a satisfactory claim, any of said query responses which claim said subject message but fail threshold quality criteria.

145. A method according to claim 144, wherein each of said query responses has associated therewith an indication of a priority level of the agent which originated said query response, and wherein said step of discarding comprises the step of discarding as a satisfactory claim, each of said query responses whose associated priority level is below a predetermined threshold.

146. A method according to claim 145, wherein each of said query responses further has associated therewith an indication of confidence with which the agent originating said query response claims said subject message.

147. A method according to claim 138, wherein said network of agents is organized as a hyperstructure of agents.

148. A method according to claim 138, further comprising the step of said network producing an output of said network in response to said step of initiating.

149. A method according to claim 148, wherein said output of said network is human perceptible.

150. A method according to claim 149, wherein said subject message includes human input.

151. A computer-implemented method for use with a network of agents including at least one agent having a plurality of downchain neighboring agents, comprising the steps of:

initiating a subject message into said network;

establishing at least a first path through said network and a second path through said network, each in dependence upon said subject message, and each at least partially distinct from the other;

and wherein at least a first agent in said first path takes local steps to produce an output of said network in response to said subject message.

152. A method according to claim 151, wherein at least a second agent in said second path takes local steps to produce an output of said network in response to said subject message.

153. A method according to claim 152, wherein said first agent is not in said second path and said second agent is not in said first path.

154. A method according to claim 151, wherein said output of said network is human perceptible.

155. A method according to claim 151, wherein said steps of establishing first and second paths comprise the steps of:

each agent receiving said subject message, which agent has exactly one downchain neighboring agent, forwarding said subject message to said exactly one downchain neighboring agent; and

each agent receiving said subject message, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said subject message and forwarding said subject message to each agent in said subset.

156. A method according to claim 155, wherein said step of determining a subset of said plurality comprises the step of determining a proper subset of said plurality.

157. A method according to claim 155, wherein said step of each agent receiving said subject message, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said subject message, comprises the step of:

each given agent receiving said subject message, which given agent has a plurality of downchain neighboring agents, consulting a prior determination, previously stored accessibly to the given agent, of the subset of said plurality which should receive said subject message from the given agent.

158. A method according to claim 155, wherein said step of each agent receiving said subject message, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said subject message, comprises the step of:

each given agent receiving said subject message, which given agent has a plurality of downchain neighboring agents, consulting prior responses from such downchain neighboring agents to a prior query inquiring whether each such downchain neighboring agent considers said subject message to be within the domain of responsibility of each such downchain agent.

159. A method according to claim 158, wherein the prior responses consulted by a first one of said agents having a plurality of downchain neighboring agents, includes non-contradictory claims of ownership of said subject message by at least second and third agents downchain of said first agent,

and wherein the subset determined by said first agent includes both said second and third agents.

160. A method according to claim 159, wherein said claim of ownership by said second agent identifies a first focus of said subject message and said claim of ownership by said third agent identifies a second focus of said subject message different from said first focus of said subject message.

161. A method according to claim 158, wherein the prior responses consulted by a first one of said agents having a plurality of downchain neighboring agents, includes contradictory claims of ownership of said subject message by at least second and third agents downchain of said first agent,

and wherein said step of each agent receiving said subject message, which agent has a plurality of downchain neighboring agents, determining a subset of said plurality which should receive said subject message, comprises the step of said first agent resolving said contradictory claims of ownership.

162. A method according to claim 161, wherein said step of resolving comprises the step of said first agent causing a user to be asked to select among said contradictory claims of ownership.

163. A computer-implemented method for processing a subject message, by a network of agents including an initiator agent and at least one agent downchain of said initiator agent, each agent in said network having a view of its own domain of responsibility, comprising the steps of said initiator agent:

preparing said subject message in response to user input;

querying at least one of the agents downchain of said initiator agent in said network whether the queried agent considers said subject message to be in the queried agent'"'"'s domain of responsibility;

resolving any conflicting responses from said queried agents, which responses meet threshold quality criteria, to identify a prevailing group of at least one of said downchain agents to whom said subject message should be passed; and

instructing at least one agent in said prevailing group to handle said subject message.

164. A method according to claim 163, wherein said step of preparing said subject message in response to user input comprises the step of preparing said subject message in response to natural language text input.

165. A computer-implemented method for processing a subject message, by a network of agents including an initiator agent and at least one agent downchain of said initiator agent, each agent in said network having a view of its own domain of responsibility, comprising the steps of said initiator agent:

querying at least one of the agents downchain of said initiator agent in said network whether the queried agent considers said subject message to be in the queried agent'"'"'s domain of responsibility;

receiving a plurality of responses in which at least two claiming agents whose responses meet threshold quality criteria each claim said subject message as being in the domain of responsibility of the respective claiming agent, said claims being in conflict;

resolving said conflicting responses from said queried agents, including causing a user to be asked to resolve said conflicting responses, to identify a prevailing group of at least one of said downchain agents to whom said subject message should be passed; and

instructing at least one agent in said prevailing group to handle said subject message.

166. A method according to claim 165, wherein said step of causing a user to be asked to resolve said conflicting responses comprises the step of asking said user to select among said claiming agents.

167. A method according to claim 165, wherein said step of resolving further includes the step of identifying said prevailing group in response to natural language text input entered in response to said step of causing a user to be asked to resolve said conflicting responses.

168. A computer-implemented method for processing a subject message, by a network of agents including an initiator agent and at least one agent downchain of said initiator agent, each agent in said network having a view of its own domain of responsibility, comprising the steps of said initiator agent:

querying at least one of the agents downchain of said initiator agent in said network whether the queried agent considers said subject message to be in the queried agent'"'"'s domain of responsibility;

receiving a plurality of responses in which at least two claiming agents each claim said subject message as being in the domain of responsibility of the respective claiming agent, each response in said plurality of responses having a respective quality level associated therewith, said at least two claims being in conflict;

resolving said conflicting responses from said queried agents to identify a prevailing group of at least one of said downchain agents to whom said subject message should be passed; and

instructing at least one agent in said prevailing group to handle said subject message, wherein said step of resolving said conflicting responses comprises the steps of;

comparing the quality levels of the responses in said plurality of responses;

evaluating how recently each of said claiming agents was instructed by said initiator to handle a prior message; and

causing a user to be asked to resolve said conflicting responses.

169. A method according to claim 168, wherein said step of resolving said conflicting responses further comprises the step of identifying said prevailing group in response to natural language text input entered in response to said step of causing a user to be asked to resolve said conflicting responses.

170. A computer-implemented method for processing a subject message, by a network of agents including an initiator agent and at least one agent downchain of said initiator agent, each agent in said network having a view of its own domain of responsibility, comprising the steps of said initiator agent:

querying at least one of the agents downchain of said initiator agent in said network whether the queried agent considers said subject message to be in the queried agent'"'"'s domain of responsibility;

resolving any conflicting responses from said queried agents, which responses meet threshold quality criteria, to identify a prevailing group of at least one of said downchain agents to whom said subject message should be passed; and

instructing at least one agent in said prevailing group to handle said subject message,

said method further comprising the step of said network, in response to said step of instructing at least one agent in said prevailing group to handle said subject message, causing an output to be generated which is human-perceptible.

171. A method according to claim 170, further comprising the step of providing said subject message in response to human input.

172. A method according to claim 170, further comprising the step of providing said subject message in response to natural language text input.

173. A computer-implemented method for processing a subject message, by a network of agents including an initiator agent and at least one agent downchain of said initiator agent, each agent in said network having a view of its own domain of responsibility, comprising the steps of said initiator agent:

querying at least one of the agents downchain of said initiator agent in said network whether the queried agent considers said subject message to be in the queried agent'"'"'s domain of responsibility;

resolving any conflicting responses from said queried agents, which responses meet threshold quality criteria, to identify a prevailing group of at least one of said downchain agents to whom said subject message should be passed; and

instructing at least one agent in said prevailing group to handle said subject message,

said method further comprising the steps of a first one of said queried agents, in response to said query;

querying at least one further agent downchain of said first agent whether the further agent considers said subject message to be in the further agent'"'"'s domain of responsibility;

receiving a plurality of further conflicting responses from said further agents each satisfying predetermined quality criteria and claiming that said subject message is within the further agent'"'"'s domain of responsibility; and

returning a response to said initiator agent claiming said subject message as being within said first agent'"'"'s domain of responsibility,

said method further comprising the steps of said first agent;

receiving said instruction from said initiator agent;

in response to said instruction, resolving said further conflicting responses to identify a further prevailing group of at least one of said further agents to whom said subject message should be passed; and

instructing at least one agent in said further prevailing group to handle said subject message, wherein said step of resolving said fturther conflicting responses comprises the step of causing a user to be asked to resolve said further conflicting responses.

174. A method according to claim 173, wherein said step of resolving said further conflicting responses further comprises the step of identifying said further prevailing group in response to natural language text input entered in response to said step of causing a user to be asked to resolve said further conflicting responses.