Information Technology Reference
In-Depth Information
moderates the impact of evidence: agents with high flexibility (i.e. great sensitiv-
ity to recent events) display more erratic behaviour than similarly flexible agents
that consult (Chap. 7 and Gooding and Addis 2004 ). As Fig. 6.1 indicates, a strong
negative bias can be overcome. However, three things are needed:
1. frequent consultation (to be made aware of new information, as illustrated in
Fig. 6.4 ),
2. a strong commitment to an alternative view held by at least some participants
(so that peer pressure alone does not force a consensus—see Fig. 6.4 , where the
actor-entropy (indifference) dips before rising steadily),
3. experimental evidence supporting an alternative view during the time of change.
The three-part inference system of abduction, deduction and induction as defined
by Peirce takes into account the possibility of change and error. For this to work,
no particular form of inference can act alone. Deduction unaided cannot deal with
the irrational sets needed to capture the phenomenology of a changing world and
of human responses to it. We have not used the abductive cycle strictly as a hy-
pothesis generator (Gooding 1996 ; Hanson 1958 ; Magnani 1998 , 2001 ). Rather,
we have emulated abduction through look-up tables representing the experimental
phenomenology for each hypothesis. These models differ inferentially and cogni-
tively, but from a functional point of view, the two methods are indistinguishable
(Addis 2000 ). For either method of implementation to work, abduction also needs
to be made part of a larger system of different, interacting inference and decision
mechanisms. In such an inference-system, the notion of 'Truth' is confined to the
internal workings of deduction. In its place, we have 'Belief'—the confidence an
individual has in statements that influence his or her actions—as a dynamic indicator
of the evaluation of a particular worldview. The notion of 'belief' as a propensity
to act displaces truth. A consequence of our approach is that actions that would be
deemed irrational according to traditional models of inference and of ontology can
be valuable in that they help agents continually test the world for change.
References
Addis TR (1985) Designing knowledge-based systems. Kogan Page, New York
Addis TR (2000) Stone soup: identifying intelligence through construction. Kybernetes 29:849-870
Addis TR, Gooding DC (1999) Learning as collective belief-revision: simulating reasoning about
disparate phenomena. proceedings AISB'99 Symposium on Scientific Creativity, University of
Edinburgh, pp 19-28
Addis T, Billinge D (2004) 'Music to our ears: a required paradigm shift in computer science'
presented at ECAP04. University of Pavia, Italy
Addis TR, Gooding DC (2004, 2008) Simulationmethods for an abductive systemin science.
MBR04: model-based reasoning. Science and engineering, abduction, visualization, and simu-
lation. University of Pavia, Italy, December, 2004. Also in Foundations of Science, March 2008.
Vol. 13, No 1, pp 37-52
Search WWH ::




Custom Search