Database Reference
In-Depth Information
sensor collection and sharing. Many sensing systems such as those in
geo-tagging applications user completely
open
frameworks in which any
participant who wishes to contribute is allowed to join the sensing envi-
ronment. This may of course result in numerous issues in terms of the
quality of the final results:
The participants who join may not be suciently trustworthy.
This may impact the quality of the results. We will formally dis-
cuss the issue of trust in a later section.
The act of participants
choosing to join
may bias the final variables
which are being tracked by the application. For example, when
the sensing is used in order to obtain feedback of a particular
type, urban participants are more likely to join because of greater
prevalence of smart phones. This kind of skew may also affect the
quality of the final results.
We note that when the recruitment is performed at the initiative of the
designer of the sensing system, a greater amount of control is achieved.
This tends to reduce the self-selection bias, which is naturally inherent in
a purely voluntary system. The work in [143] observes that the process
of recruiting volunteers for participatory sensing campaigns is analogous
to recruiting volunteers or employees in non-virtual environments. This
similarity is used in order to create a three stage process for recruitment:
Qualifier:
This refers to the fact that the participants must meet
minimum requirements such as availability and reputation. This
ensures that high quality responses are received from the partici-
pants.
Assessment:
Once participants that meet minimum requirements
are found, the recruitment system then determines which candi-
dates are most appropriate based on both diversity and coverage.
This ensures that bias is avoided during the recruitment process.
Progress Review:
Once the sensing process starts, the recruit-
ment system must check participants' coverage and data collection
reputation to determine if they are consistent with their base pro-
file. This check can occur periodically, and if the similarity of
profiles is below a threshold, this is used as a feedback to an addi-
tional recruitment process.
We note that the above process is quite similar to that of an employee
hiring process in an organization, which is designed to maximize diver-
sity, reduce bias, and maximize quality of the volunteers.
