Summary
The purpose of this project was to develop and execute improved research methodology for studying how consumer information privacy perceptions and behaviors change over time. This project is unique because most of the behavioral research regarding information privacy (and with mobile devices in particular) had previously been based entirely on surveys and laboratory experiments with low external validity. Therefore, to accomplish our objective, several mobile applications were developed or improved with built-in capabilities for experimental manipulations which were tested in real-life field studies. We found several interesting new findings which have resulted in published conference paper proceedings with student authors with journal papers on the way.
Key Finding
The findings are best understood within the mobile applications developed to test our theory. The first application is a geo-caching game which incorporates an online social network. This game was well-received by participants who reported high enjoyment and engagement with the app. The app, named “Findamine” incorporated several types of consumer information including GPS-based location data, personal information included in their game social network profile, and social data including friends tracked and followed electronically during the game. In other words, there was a significant amount of potential information risk. With IRB approval, we withheld informed consent so that participants would perceive real information risk because they believed the app was developed by an external company (rather than academic researchers). Because of this unique design, we discovered several interesting findings.
First, consumers are not “rational” (in the academic sense) when making disclosure decisions. This does not mean that consumers simply disclose “too much.” Rather, it means that consumer behavior does not typically follow a rational “cost/benefit tradeoff” (Vroom, 1964). In other words, when the benefits of information disclosure increase, consumers do not always disclose more information. Similarly, when the benefits of disclosure decrease, consumers do not necessarily decrease their disclosure. In fact, most consumers behave the opposite from this expectation. To prove this, we manipulated the Findamine game to incentivize profile information disclosure by offering a certain number of points for each piece of information (e.g. name, address, phone, relationship status, income, education, birth date, etc.) they entered into the system which changed over the 8-12 weeks of the game. These points increased the likelihood that they would win some of the weekly gift cards or the end of game tablets that were given out.
Participants were randomly assigned to one of two treatments. Either, 1) the points per profile disclosure increased gradually over time, or 2) the points decreased at the same rate over time. The results revealed that most of those in the increasing points condition actually decreased their disclosure over time and removed some of their personal data. On the other hand, most of those in the decreasing points condition actually increased their disclosure over time.
This result may seem perplexing at first. However, a very logical explanation can be found in prospect theory (Kahneman & Tversky, 1979) which explains some of our human irrationalities. In particular, we behave irrationally when we have imperfect information and because our level “risk aversion” can easily change. This theory explains the irrational behavior of gamblers. For example, a gambler who begins with $1000 and wins another $1000 is in a “gain” position relative to where they began (a.k.a. their “reference point”). Once in a gain position, the gambler will decide to take fewer risks—even though the actual likelihood that they will win or lose has not changed. Similarly, a gambler who has lost half of their money will start to take greater risks (because they want to return to their original reference point); again, even though the risk hasn’t changed.
The same behavior was exhibited in our study because consumers are unable to assess the true risk of disclosing information (i.e. they have no idea what Findamine intends to do with their information), they make use of a reference point to help them make decisions. For example, the participants’ initial disclosure decision when they registered to play the game was based on some perception of their current overall risk profile. Once the disclosure points began to change, participants based their subsequent decisions on how far they deviated from that initial cost/benefit tradeoff. For example, if I suddenly get more points for information I’ve already disclosed in the game, I perceive myself to be in a “gain” position relative to where I began. Therefore, I become more risk averse and I actually remove some of the data I entered, which gets me closer to my original reference point, or risk/benefit balance. Similarly, those who lost points over time because to disclose more information in order to back to their original point level.
This finding is under preparation for submission to our field’s top journal (MIS Quarterly). However, several other interesting findings have already been published in conference proceedings and are also under preparation for journals. The next section outlines each publication and the roles each student played in the work.
Privacy Fatigue
From the same experiment above, we also manipulated the features available in the privacy control settings available to each participant. In particular, each player was randomly assigned to have privacy settings options that were either 1) very simple, 2) moderately complex, or 3) very complex. We found that when initially playing the game, those with very complex privacy settings were the most satisfied and excited about the features. However, after playing for several months, they were dissatisfied with how difficult the complex controls were to use in practice. As a result, they ended up disclosing more information overall if we set privacy defaults to share the maximum amount of data. On the other hand, those with very simple controls were the least excited at first, but the most satisfied over time and share the least data. The result is the concept we refer to as “privacy fatigue.” A conference version of this paper was published in the proceedings of our discipline’s most prestigious conference and was nominated for best paper award:
Keith, M. J., Maynes, C. Lowry, P. B., and Babb, J. S. (2014). Privacy fatigue: The effects of privacy control complexity on consumer self-disclosure. Proceedings of the International Conference on Information Systems (ICIS), December 14-17, Auckland, NZ
Courtenay Mayes was the graduate student (2nd author) who helped on this paper. She and I have collected a new round of data over the last year and are preparing to submit this paper to MIS Quarterly. It received excellent feedback at the conference and one of the senior editors suggested we submit the paper to that journal. Much of the MEG funding went toward Courtenay’s wages and travel funding for both of us.
Self-Regulation
Another finding from this line of research is that consumer self-regulation (a.k.a. “patience”) plays a much larger role in determining their online information disclosure than their actual perception of privacy risk. Nam Ngo was the graduate student who worked with me to create another app called “Sharing Tree” which was used to demonstrate this finding. We published two conference papers on this project and are preparing a journal paper version for the European Journal of Information Systems.
Keith, M., Ngo, N. and Babb, J. (2014). The effects of consumer self-regulation on mobile information disclosure. Proceedings of the Americas Conference on Information Systems (AMCIS), August 7-10, Savannah, GA.
Keith, M. Ngo, N., and Babb, J. (2014) The effects of consumer self-regulation and risk immediacy on mobile information disclosure over time. Proceedings from the Dewald Roode Workshop on Information Security (IFIP WG8.11/11.13) 2014. Newcastle, UK, June 16-17.
Nam Ngo was unable to attend either conference with me. However, much of the funding also went to his wages and my travel to present at these conferences. Interestingly, I have had offers from colleagues at other universities (who attended these presentations) to have Nam apply for their doctoral programs. Nam is still considering his options but would be welcomed at a number of prestigious programs as a result of this work.
Brand Influence
Lastly, this MEG grant also funded my work with an undergraduate student (now a graduate student) who was interested in how perceptions of brand credibility and recognition affects consumer’s initial risk perceptions when deciding whether or not to download a mobile app. Thong Pham has helped me execute another data collection and experiment also using the Sharing Tree app. Our data shows that brand perceptions heavily outweigh risk perceptions when making disclosure decisions. We are currently preparing a manuscript directly for journal submission to the European Journal of Information Systems from this project. Some of the MEG funding was also spent on Thong’s student wages
Mentoring Environment
Overall, I am very pleased with the result of the mentoring environment with these students. I have chaired dissertations (or been a committee member) at other universities and I believe these three master’s students are more capable than other doctoral students I’ve worked with. We had weekly meetings and all three students participated in every facet of the project from idea generation to theory development to data collection and even manuscript preparation.
Budget Allocation
The MEG budget was spent almost exactly as it was anticipated. The amount for the actual dolalrs spent is greater than the initial proposal of $20,000 because it also includes funding from my own 20 account.
Purpose | Proposed | Actual |
Undergraduate wages | $1800 | $4230 |
Graduate wages | $8400 | $8100 |
Supplies (prizes for participants) | $2835 | $4955 |
Travel (to UK, New Zealand, GA) | $6965 | $6965 |
Total | $20000 | $24250 |
Conclusion
In conclusion, this MEG funding has provided a great mentoring and research experience for both myself and these students. We have developed strong relationships that will continue as they enter their own PhD programs. With the publications thus far (including the best paper nomination at our most selective international conference) and those on the way, I believe our results indicate that this has been “money well spent” by the university. I hope to continue this level of mentoring and productivity in the future. Thank you for the opportunity!
References
Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica, 47(2), 263-291.
Vroom, V. H. (1964). Work and Motivation. New York: Wiley.