Information Technology Reference
In-Depth Information
-1500
-1000
-500
0
500
1000
1500
2000
RANDOM
Costs_TRUST
TRUST_Ambient
Low_COST
FIRST_TRUST
STAT
RANDOM
Costs_TRUST
TRUST_Ambient
Low_COST
FIRST_TRUST
STAT
RANDOM
Costs_TRUST
TRUST_Ambient
Low_COST
FIRST_TRUST
STAT
Figure 11.16
Experiment 8: comparing different delegation strategies in different cases: number of
tasks performed (success rates, case 1), number of credits (gains, case 2), and number of analyzed offers
(time spent, case 3). (Reproduced with kind permission of Springer Science+Business Media
C
2005)
In
EXP9
(see Figure 11.17) we use controllability and we offer the possibility of re-posting
a task. An agent who uses controllability as an additional parameter for trust (i.e. giving a non
null weight to the corresponding edge in its FCM) is compared with agents who do not use it; it
is (more or less) biased towards choosing agents who give more reports and so they have more
possibilities of re-posting unsuccessful tasks (tasks can be reposted only once). All agents (ex-
cept
TRUST BASE
) use the following heuristic in order to decide whether to retire delegation:
if
(10
−
current essay
/
current essay
<
1)
and
(
success number
<
3)
and
(
I can repost
)
then retire delegation
The experiment was performed for 200 tasks and 250 simulations.
RANDOM
is the baseline;
TRUST
is the agent who uses ability, willingness and environment but without controllability;
TRUST BASE
is the same agent but the one who never reposts their tasks;
CONTROL 1
Search WWH ::
Custom Search