Information Technology Reference
In-Depth Information
Table 7.10
Belief system 2.0—narrative for cycles 0-10 in Fig.
7.4
Cycle 1: Solo—Consultation Solo 0.606287 sphere
Cycle 2: Solo—single_image
Make_observation single_image
circle
0.606287 sphere
Cycle 3: Solo—No_Experiment 0.788873 sphere
Cycle 4: Solo—No_Experiment 0.788873 sphere
Cycle 5: Solo—No_Experiment 0.788873 sphere
Cycle 6: Solo—No_Experiment 0.788873 sphere
Cycle 7: Solo—No_Experiment 0.788873 sphere
Cycle 8: Solo—Consultation Solo 0.788873 sphere
Cycle 9: Solo—single_image
Make_observation single_image
point_circle
0.788873 sphere
Cycle 10: Solo—No_Experiment 1.000000 cone
effects. Figure
7.5
shows how the confidence of two agents in each of the three
hypotheses changes, as consultations and new observations are made. Actor 1 starts
with negative bias (see Fig.
7.5
) while actor 2 is unbiased. They are otherwise identical
as to flexibility, receptivity and access conditions.
Consultation has several effects:
• First, comparing actor 1 in Figs.
7.4
and
7.5
shows a jagged but less erratic
trajectory.
• Second, the actors converge to the same conclusion. Multiple runs of this scenario
show that the sequence of events producing agreement varies in each run (compare
the two narratives in Table
7.11
,
which also show that self-consultation occurs
less frequently than for a solo actor).
• Third, consultation can also moderate the effects of bias.
To understand the macro-behaviour displayed in the plots we would need to examine
the actions of each actor and their outcomes in each of their iterations in more detail.
7.7
Conclusions
We have shown in this chapter how simulation methods enable us to model the
interaction of hypotheses; experiments and an emerging phenomenology and that
iterative modelling can represent the combined influence of evidence and opinion
on inferences about hypotheses. Studies made with the belief-revision system show
some behaviour that is stable over variation of parameters. These include convergence
of beliefs to the objective hypothesis, that an emerging consensus is often interrupted
and then restored, and that an optimum balance between the impact of evidence and
the influence of opinions is achieved for an event-window of 2-3 events.
