2.4 Chapter Summary
3.1.1 Participatory sensing
The prevalence and connectivity of sensor-rich smartphones extends their capabilities far beyond the individual level, into something of a global sensor network, through which sensing applications can collect large quantities of data. In participatory sensing studies, participants are often individual data sources within a larger project that aims to answer research questions on a community, national, or even global scale. In some cases, a participant simply has to download and run an app, which autonomously collects relevant data. The simplicity of contribution to the altruistic goals of these large-scale studies, has participants acting as“citizen scientists”, often supplying data for a collective good rather than their own personal benefit [68].
As opposed to traditional ESM studies in psychology, the focus of a participatory sensing study is on an individual’s environment, rather than the individual themselves. This is the primary difference between the two types of study, such that participatory sensing studies are often implemented for studying environmental or sociological issues. The development of sophisticated sensor capabilities enable participants to have different levels of involvement in the data collection process, defined in terms of two types of “crowdsensing” as follows:
• Participatory crowdsensinghas participants decide the appropriate conditions under which to engage in a study. Although sensor data may be acquired, it is primarily obtained through self-report surveys, and thus requires participants’ active, conscious involvement in a crowdsensing application.
• Opportunistic crowdsensingapplications run in the background, choosing when to passively collect data, such that participants can be unconscious of any sensing that takes place. As an example, complex classifications of participants’ mobile sensor data was used to predict the expected arrival of buses [69].
In general, participatory sensing studies are more flexible than ESM studies in order to minimise burden on participants, given that they do not necessarily receive financial compensation for their participation. For example, time-contingent sampling is seldom employed, instead relying on participants to engage in manual event-contingent sampling. Further, additional incentives must be provided to the participant in lieu of compensation. A recent review of non-monetary incentive mechanics in health-based participatory sensing apps was conducted by Anawar et al [70], in which self-monitoring features were found to be the most reliable in maintaining
3.1. SMARTPHONE ESM DEVELOPMENTS 37
Table 3.1:Participatory sensing incentives and their applicability to ESM studies
Incentive Definition Applicability to ESM
Monetary Participants rewarded with money Current practice. Not sustainable for large sample sizes
Collective Participants motivated by
attainment of a ‘common good’
Not useful. Purpose of ESM study unlikely to be motivated by community benefit
Social (interaction)
Social networks leveraged, allowing participants to share and interact
Not useful. Privacy issues prevent
sharing and interaction with other participants
Social
(self-interest)
Participants receive collective info
through contributing: ‘quid pro quo’ As above
Monitoring Participants empowered to monitor
and manage their own data
Potentially useful, participants could be incentivised by monitoring study progress
Autonomy Participants in full control of what
they send, where and when
Potentially useful, if scope for flexibility in data collection process is feasible
Feedback Participants can send and receive
feedback on their progress
Potentially useful, ensuring feedback both ways is kept private and anonymous
engagement. This result is consistent with behaviour theories that support individuals’ perceived autonomy to make choices and exert influence as a motivational factor [71, 72].
In their application to understanding social and environmental conditions, participatory sensing apps are often employed in fields of research such as social science, where researchers are less likely to have significant programming experience. As such, tools to allow creation of participatory sensing apps are often a necessary prerequisite for such research.
3.1.1.1 Authoring Tools
A number of existing tools allow researchers to orchestrate “participatory sensing campaigns”. EpiCollect [73] and Open Data Kit [74], for example, are tools for designing questionnaire forms that can be deployed onto participating users’ devices. They allow a variety of question types, and multimedia data inputs (such as audio, video and photo) with optional branching and skip logic. Researchers and participants can view their collected data in the form of graphs and charts on a project website. However, such applications still rely on participants’ incentive to initiate the surveys themselves, limiting any form of ESM applications to event-contingent sampling schedules. Furthermore, the complexity and diversity of sensor classifications would make it impossible to allow non-programmers to orchestrateopportunisticcrowdsensing apps.
38 CHAPTER 3. SMARTPHONE DATA COLLECTION
Nevertheless, participatory sensing is still of relevance to the work of this thesis. Given the burdensome nature of experience sampling studies, and the financial costs to the researcher in scaling up to larger populations, participatory sensing research describes additional means of minimising participant burden and improving experience that could tentatively be incorporated into appropriate ESM studies. In particular, self-monitoring incentive mechanisms appear most appropriate for application to experience sampling: allowing participants to monitor their collected data, ask questions and receive feedback, and tailor their data collection preferences, could motivate compliance. Table 3.1 thus summarises the various types of incentive that have been applied in participatory sensing studies, and their applicability to ESM studies.