Methods in Human Computer Interaction/Quantitative/An investigation into developing trust in the rideshare industry
Introduction and Research Question
Ridesharing services and their concomitant mobile apps have taken the transportation-for-hire industry by storm over the last three years (Ferenstein, 2014). This incredible expansion of the rideshare industry, and its possible effect on the U.S. economy, has even been discussed in revolutionary terms (Ciaccia, 2014). But this rapid growth has not been without controversy. There have been a number of notorious incidents involving rideshare drivers sexually assaulting passengers (Hussain, 2015; Tempera, 2014). Rideshare drivers have also been the cause of several high profile accidents resulting in passenger or pedestrian deaths (Vara, 2014). This often felonious behavior has engendered public mistrust in the nascent rideshare industry. Current and prospective rideshare customers have become concerned for their safety, while at the same time wondering whether it is rideshare drivers or the rideshare industry that are liable for damages when something like this occurs. The two largest rideshare services, Lyft and Uber, have responded by fighting the passage of any new regulations that might make them guarantee their drivers are appropriately insured (Lieber, 2014).
More than any other threat to the emergent rideshare industry’s continued growth, it is the lack of user trust that can have the most dire consequences. Trust, it has been said, is the one thing that makes all sharing economy businesses work (Fung, 2014; Tonkinwise, 2012). Möhlmann describes trust as a user’s “faith in a provider’s reliability, and the impression of security during use” (2015). But how do rideshare companies create and build trust among its users? It is the goal of our project to create a viable solution to this problem. We investigate how best to communicate to end users the safeguards that are in place for riders. More specifically, we explore the content, size, style, timing and placement of the information we propose to be shared with rideshare users (Egelman, Tsai, Cranor, & Acquisti, 2009; Kelley, Cranor, & Sadeh, 2013; Lin et al., 2012).
If driver liability coverage, and other passenger protections, were more clearly communicated within rideshare apps, would trust in ridesharing services be increased?
User Community, Sample/Population
Vehicle-for-hire mobile applications, otherwise known as rideshare, like Uber, Lyft, Sidecar and others, are prone to use by certain demographics, simply because of the technology required to use them. With that in mind, our proposed user community can be narrowed to some easily discernible categories. Previous research in the area helps us define this user community.
In one recent study, the University of California Transportation Center discovered some reinforcing facts about the rideshare-app-using community, specifically. The typical app user, as you can see, is young (between 25 and 44) and middle income to affluent ( making anywhere between $30K and $200K primarily). In the area of income, the largest number of users fit into the affluent category, making between $100K and $200K/yr. Respondents were also typically well educated. 84% of the survey respondents had a bachelor's degree, compared to just 54% of all San Fransicans over age 25.
A similar study was conducted in the Seattle area, to gauge use of taxis and other car-for-hire services. Demographically, the study provided similar results. Users of rideshare application services were young (between 21 and 44 primarily) and middle-income to affluent (the majority making $60K/yr or more).
Marketing messages, and branding, are another key to identifying the target user and likely distribution of the user community. As you can see, each of the major rideshare/vehicle-for hire services, branding and marketing imagery are targeted at audiences identified in the studies we've just referenced.
Noting the various research on the subject, as well as prevailing messaging and branding, the sample population for our study design would be emblematic of the user community. We would seek respondents who are:
- Primarily early to middle aged adults (25 - 45)
- Middle to high income ($40 - $200K)
- Educated (high school graduate or higher)
The sample size for the study is more difficult to properly identify. Without a true number of users (due to a lack of a published number of users), determining the exact number of necessary responses for a significant sample is only an estimate. However, an estimated target number is better than none.
If we estimate a rideshare/on-demand vehicle user base of at least one million (probably very conservative), we would need 600 respondents for a significant sample that could deliver a 95% confidence level and a margin of error of 4%. This would be the initial target.
As we begin to look at methods, we also take a look at our independent and dependent variables that will shape our choices and research.
The object of this research is to determine the level of trust that users have in vehicle-for-hire mobile applications. In order to do so, our survey assesses the trust level initially, then presents differing circumstances that could potentially affect user trust. Thus our variables are defined as follows:
Our Independent Variable is the Presence of specific messaging i.e. insurance info, background checks. We present respondents with different information scenarios that may affect their levels of trust. The Dependent Variable is the user's Level of trust, which we would hypothesize might change based on the information presented in the application
We measure trust as a form of sentiment, on the Likert scale of the responses. Users will respond on a 1-5 scale, at level of agreement, with trust being the subject of the question. The highest level of trust would be 5.
Taking a holistic look at both qualitative and quantitative methodologies which could be used for this type of research study, we outline three methodologies in this section; focus groups, interviews and surveys. We will recommend one of these methodologies for our research study. One thing to keep in mind is the way in which trust is attained can be the key communication factor at various touch points within the user experience of the ride sharing app. Ride sharing is a relatively new type of service and it is important to communicate that trust effectively early on. The way in which that trust is attained is key communication at various touch points within the user experience of the ride sharing app.
Focus groups consist of usually 10 participants or less. These participants would discuss a specific product, service or topic. Participants can be asked questions with follow up questions from the researcher and feedback from others within the group. Participants can also be asked to complete a specific sentence about ride sharing and trust which will help them think through their responses instead of answering broad questions.
Some drawbacks of focus groups is they take longer, have a higher cost and require a skilled researcher. In addition, because trust issues can lead to answers about personal experience, participants may be reluctant to delve into the details of their experiences in a group setting.
Interviews can be conducted in a number of setting types where their purpose is to collect and exchange data. They are adaptable in that they can be structured, semi-structured or unstructured to gather further insight. Participants can also demonstrate their experiences for ride sharing whether favorable or not. The use of artifacts or probes can help facilitate these communications to further understand the picture the participant is giving from their experiences. Since interviews are one on one, users may be more open to sharing their experiences as opposed to a focus group for complex and sensitive topics such as ride sharing and trust.
As with focus groups, some drawbacks to interviews are that they can be time consuming, with a higher cost and require a skilled researcher. Also, participants may be reluctant to answer honestly in person as opposed to being anonymous.
For our recommended methodology, we suggest a quantitative approach using a survey as the instrument. Surveys are a quick, cost effective way for gathering information. They typically consist of a set of questions used to assess a participant’s preferences, attitudes and opinions toward a product, service or topic. A key advantage is that a survey can be administered through the app during peak hours of ride sharing use to gather a higher level of data and feedback. From our sample population, we can compare demographic data to the results of the survey. Data captured along with analysis of the demographic data, can not only be applied to improve the app for communicating user safety and trust, but can aid in the design of other methodologies we may want to use later on.
The analysis of the Likert Scale data begins with a decision on how to view the data - Interval or Ordered, Categorical. Viewing the Likert Scale data through the Interval lens amounts to believing that values can exist between the five options. For example, there could technically be an average response that is slightly more than “neither agree nor disagree”, but is not quite “agree”. However, for the purpose of this study it was decided that the data should be viewed through the Ordered, Categorical lens, where an answer can only be one of the five responses. This removes any calculations involving mean.
Once enough participants had completed the survey the analysis could begin. The first step would be to code the captured data. For the Likert Items, the coding would be as follows:
1 - Strongly Disagree, 2 - Somewhat Disagree, 3 - Neither Agree or Disagree, 4 - Somewhat Agree, 5 - Strongly Agree
For the demographic data the coding would be as follows:
Gender: 1 - Female, 2 - Male
Age: 1 - 18-24, 2 - 25-34, 3 - 35-44, 4 - 45-54, 5 - >55
Household Income: 1 - $0-$50,000, 2 - $50,001-$100,000, 3 - $100,001-$150,000, 4 - $150,001-$200,000, 5 - >$200,000
Descriptive statistics would be helpful to get a basic feeling for the data. Useful information that could be captured includes the median, mode, and quartiles. Additionally, a chi-square test could be performed to analyze the frequency counts or a Mann-Whitney test could be run if the assumptions are not met.
Finally, various charts and graphs could be created and would provide a useful visualization of the captured data.
As with all subject pools and proposed research, it’s endlessly important to keep your subjects in mind both ethically and quantifiably, especially when it comes to such delicate matters like possible assault, trust, and overall feelings of safety. Because of this, we had to have lengthy debates about what kind of data we felt best suited our research goals but also granting a voice and understanding of the people affected or not affected. Ultimately, this lead us to design a quantitative survey providing us with a platform to better inform the greater understanding as well as a basis for future, qualitative studies later on. As stated before, we attempted to keep a consistent understanding of the potential gravity of the situation for some subjects when working on the survey. And made efforts to evaluate what impacts, if any, it could have on potential answers when designing questions.
During our design, we considered three main strategies for gathering data. Understanding the context in which we decided between these three will provide an understanding of the pros and cons of each, and ultimately the reasoning behind our decision of one. First were focus groups, which can play to strengths of security in numbers as well as jogging the memory of shared or similar experiences. Informational recall can sometimes be difficult with ridesharing services as I might point out, as often people are more concerned with the destination and not the journey so to speak. In terms of negative aspects, focus groups especially about sensitive material like this, could hurt responses as people could fear looking vulnerable in a group of strangers. Due to issues with focus groups, we began looking at interviews which could help us understand participants while also giving them a more personal experience ideally free from feelings of judgment. Interviews are not without issue however, as effects like the sheer amount of hours required to conduct would limit the amount of people we could get to and experiences heard. Something very important with an industry the size of ridesharing that’s only growing in popularity. With the other two options eliminated for now, we ultimately decided on survey questionnaires as they would allow us to gather a large pool of data, from which we could consider pulling specific individuals later. And also notice common trends that could easily inform our continued research later on. The obvious limitation would be lack of complex qualitative context, but we felt for this initial study it would be worthwhile to gain a quantitative understanding first then a more tailored understanding later.
These are two quotes from related literature that would inform our understanding of survey design as well as our choice to go with questionnaires.
- "The study found that most women who completed a questionnaire including sensitive items about early childhood and adult forms of victimization generally found the experience to be a positive one." (E. A. Walker et al., 1997)
- "We noted that for every offense but one, survey means were greater than those of the official data." (Decker, 1977)
The first is a published article about the efficacy of questionaries’ about assault on women which found that the research wouldn’t impact most female participants negatively. There are, however, important implications on gender and this type of subject matter that were not touched upon in the article but are extremely important nonetheless. For example, men’s truthfulness of assault in a focus group might be compromised due to societal pressure. The second article from the Journal of Criminal Justice found that survey results were actually reportedly higher than official data on crime. Proving that official data of almost all violent crime was not only not accurate, but that people might not even be reporting it unless sought out. This is important because it shows the impact that surveys could have on the industry if users simply aren't reporting crimes they otherwise might disclose on a questionnaire. After all, a crime may not be reported but the damage of trust would still very much exist.
Ciaccia, C. (2014, November 14). How Uber Is Changing the U.S. Economy Like Never Imagined. TheStreet. Retrieved from http://www.thestreet.com/story/12954695/1/how-uber-is-changing-the-us-economy-like-no-one-ever-imagined.html
Decker, S. (1977). Official Crime Rates and Victim Surveys: An Empirical Comparison. Journal of Criminal Justice, 5, 47-54.
Egelman, S., Tsai, J., Cranor, L. F., & Acquisti, A. (2009). Timing is Everything?: The Effects of Timing and Placement of Online Privacy Indicators. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 319–328). New York, NY, USA: ACM. http://doi.org/10.1145/1518701.1518752
Ferenstein, G. (2014, September 18). The rise of Uber and the demise of taxis, in one chart | VentureBeat | Business | by Gregory Ferenstein. VentureBeat. Retrieved from http://venturebeat.com/2014/09/18/the-rise-of-uber-and-the-demise-of-taxis-in-one-chart/
Fowler, FJ (Jr), Mangione, TW (1990) Standardised Survey Interviewing. Sage Publications, Newbury Park. http://www.usabilitynet.org/tools/interviews.htm
Fung, T. (2014, November 10). Why reputation is the currency of the future. Financial Review. Retrieved from http://afr.com/technology/web/ecommerce/whyreputation-is-the-currency-of-the-future-20141110-11jpr2
Hussain, R. (2015, January 14). Another Uber X driver charged with sexually assaulting customer. Chicago Sun Times. Retrieved from http://chicago.suntimes.com/news-chicago/7/71/291426/uber-driver-charged-sexually-assaulting-customer
Kelley, P. G., Cranor, L. F., & Sadeh, N. (2013). Privacy As Part of the App Decision-making Process. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 3393–3402). New York, NY, USA: ACM. http://doi.org/10.1145/2470654.2466466
Lazar, Johnathan. (2010, February 1). Research Methods in Human-Computer Interaction.
Lieber, R. (2014, September 5). The Question of Coverage for Ride Service Drivers. The New York Times. Retrieved from http://www.nytimes.com/2014/09/06/your-money/auto-insurance/offloading-the-risk-in-renting-a-car-ride.html
Lin, J., Amini, S., Hong, J. I., Sadeh, N., Lindqvist, J., & Zhang, J. (2012). Expectation and Purpose: Understanding Users’ Mental Models of Mobile App Privacy Through Crowdsourcing. In Proceedings of the 2012 ACM Conference on Ubiquitous Computing (pp. 501–510). New York, NY, USA: ACM. http://doi.org/10.1145/2370216.2370290
Möhlmann, M. (2015). Collaborative consumption: determinants of satisfaction and the likelihood of using a sharing economy option again. Journal of Consumer Behaviour, n/a–n/a. http://doi.org/10.1002/cb.1512
Oppenheim, AN: Questionnaire design, interviewing, and attitude measurement (New Edition). Continuum press: London and New York, 1992. http://www.usabilitynet.org/tools/surveys.htm
Rayle, Lisa, et al. "App-Based, On-Demand Ride Services: Comparing Taxi and Ridesourcing Trips and User Characteristics in San Francisco University of California Transportation Center (UCTC)." (2014). http://www.uctc.net/research/papers/UCTC-FR-2014-08.pdf
Saran, S, Ke, K. How to Use the Likert Scale in Statistical Analysis. Statistics Cafe. May 2011. Retrieved from <http://statisticscafe.blogspot.com/2011/05/how-to-use-likert-scale-in-statistical.html> Apr 2015.
Tempera, J. (2014, December 18). Uber driver is accused of rape, kidnapping of customer - The Boston Globe. The Boston Globe. Retrieved from https://www.bostonglobe.com/metro/2014/12/17/uber-driver-accused-rape-kidnapping-customer/KYnOQczKFqggfbnri2FpGL/story.html
Tennessee Transportation and Logistics Foundation, Taxi Research Partners. City of Seattle and King County Taxi, For Hire Vehicle and Limousine Services Demand Study. Seattle, WA: City of Seattle, 2013. http://www.seattle.gov/council/issues/taxis/13100301JC%20Seattle%20Final%20Report%20Master.pdf
Tonkinwise, C. (2012). Sharing Trust: Tasteful Designs of Social Systems. In Trust Design Publication 4: Public Trust Trust Design. Premsela Foundation. Retrieved from http://trustdesign.org/?p=89
Vanek, C. Likert Scale - What is it? When to Use it? How to Analyze it? SurveyGizmo.com. April 24. 2012. Retrieved from <http://www.surveygizmo.com/survey-blog/likert-scale-what-is-it-how-to-analyze-it-and-when-to-use-it/> Apr 2015.
Vara, V. (2014, November 4). Uber, Lyft, and Liability. The New Yorker. Retrieved from http://www.newyorker.com/business/currency/uber-lyft-liability
Walker, E., Newman, E., Koss, M., & Bernstein, D. (1997). Does the Study of Victimization Revictimize the Victims. General Hospital Psychiatry, 19, 403-410.
Writing, Alexis. (2009). Advantages & Disadvantages of a Focus Group. Houston Chronicle http://smallbusiness.chron.com/advantages-disadvantages-focus-group-784.html