In this project we are exploring the viability of what we call the National Experience Sampling Project (NESP). This would be a nationwide, opt-in, population scale program where ordinary users of mobile phones help researchers gather data for medical research and improving public health. What would people contribute? Just a few idle moments here and there and some information that might be gathered by their mobile phones. The NESP project would take advantage of context-aware experience sampling, where mobile phones might automatically determine appropriate times to ask context-sensitive questions. For example, a phone that can detect if someone is walking may permit a researcher to ask a question about motivations for getting exercise only after what was just detected as an unusually long walk for one individual.
Why is something like NESP necessary?
Researchers actually have a rather poor understanding of how everyday life impacts overall health. It is very hard to ask someone to recall things like exactly what they ate, how much television they watched, how many minutes they were physical activity, where they may have been exposed to environmental hazards, how stressed out they feel, and other things that might impact their long-term health. Today researchers primarily use paper surveys to gather this information, which can be grueling and error prone. This is one reason that the health “advice” that you get about what to do, or not to do, changes so often. Researchers in behavioral medicine are all in agreement that better measurement tools are needed. Computer technologies – especially mobile computer technologies like phones – may strengthen the quality of research.
Another problem with traditional methods of gathering information about what people do and how that activity relates to health is that it is expensive to administer surveys and process the results. Most studies, therefore, involve small groups of people and either ask people questions very infrequently about long periods of time (e.g., “Over the last year, rate your average stress level”) or they ask people to keep diaries of how they feel for very short periods of time (e.g., “Record any interaction that substantially increases your level of stress for the next week”). Consequently, it is difficult to study how the ebb and flow of everyday life impacts how people feel and what they do. Take stress, for example; researchers would still like to better understand how stress levels change over time and why, and then how that may or may not impact behaviors such as eating and sleep.
The bottom line is that researchers need better tools to gather information on what people do as they do it, and they need to gather that information from many thousands of people for months or years continuously. We think that mobile phones offer an unusual opportunity to lower the cost and improve the quality of some types of data collection, possibly permitting new types of research studies and analyses that could lead to breakthroughs in science and medicine.
To illustrate the potential of mobile phones to provide data for health research, consider a scenario where health researchers are interested in measuring psychosocial stress. By 2011, mid-range phones will have all the capabilities of the best current devices plus more memory, more processing power, knowledge of location (GPS or GSM based positioning), built in 1-3 axis accelerometers, higher-quality cameras, faster data transfer speeds, and better low-power wireless communication capabilities (for collecting data from on-body sensors). These capabilities are being added to enhance phones’ functionality for the purposes of communication, entertainment, and time management. The phones will be able to infer information such as where someone is, if and how that person is traveling, and if he or she is getting exercise.
Suppose that when the user of such a phone signed up with the mobile service provider, the phone buyer could “opt-in” to a voluntary public health service called the National Experience Sampling Project (NESP), or perhaps the National Public Health Experience Project (BPHEP) or some other name. Participants in the NESP simply agree to receive up to three quick multiple-choice questionnaires a month on their phone that may help researchers improve public health. The NESP is for a good cause, endorsed by credible national health organizations, and costs nothing, and participation does not take much time or effort, so the phone buyer signs up. The program is similar to popular distributed computing projects such as the Great Internet Mersenne Prime Search, SETI@Home, or the Human Proteome Folding Project, but instead of donating home computer “idle” time, people donate a bit of their own “idle” time to answer questions “for the sake of health and science.” The phone buyer may also sign up to help the researchers test novel health interventions, to share their raw sensor data, and/or to receive queries at a higher rate, perhaps in return for the opportunity to participate in a lottery for small prizes. The introductory materials explain the mechanisms that are in place to protect the phone user’s privacy. Participation is free because the mobile phone provider donates the small amount of digital bandwidth required to send the NESP data as a public service, much like traditional media venues donate airtime for public service announcements. Alternatively, the researcher pays for the transfer of an inconsequential amount of encoded data.
Once the phone is functional, the NESP participant will start receiving questions: the phone rings or vibrates and a multiple-choice question appears on the screen in a large font. The graphical user interface is designed to be easy to use. The question can be ignored if it is received at a bad time, but most interactions are fast, and questions only appear a few times a month. The first set of questions the person is asked request basic demographic information. Sometime after that, the participant will receive a variety of questions (often triggered by a particular context), such as, “Rate your level of stress after that lengthy phone call,” or “Were you taking a smoke break?”
Study Proposal Submission Process
Meanwhile, health researchers are continually submitting proposals to a NESP review board, which consists of prominent health researchers. Proposals are competitively reviewed; each must justify the tradeoff between potential research information gained versus “cost” to the end user in intrusiveness and time. For example, some proposals may only require sending a single, short set of questions to a subset of the NESP user base. Others may involve sending a set of questions to the entire NESP user base at the same exact time. Some may use questions that change based on previous answers or on personal baselines of NESP users, which may be acquired automatically by sensors on the phones or other devices created for special research projects, such as the NIH’s new Exposure Biology Program. Others may ask questions only when a NESP participant is in a particular situation, as determined automatically by the phone. For example, a researcher could have the phone ask a question about physical activity attitudes or stress levels only when someone has been walking for some minimum amount of time (e.g., 50% longer than that person usually walks) and is in a particular type of space (e.g., large park), as determined by the accelerometers, GPS, and a GIS database. The typical NESP participant receives a small number of interruptions per month. Some interactions could be structured as computer adaptive tests84, so that a measure can be established with only a small number (< 10 and usually 1-2) of quick questions. Other interactions could be simulcast: a message could be sent to 100,000 people who are probably sedentary (based on phone sensors) asking if they are watching television right now. Those who answer might then be asked if some item in the news has contributed to their perceived stress levels. Rapid survey generation and questioning could be conducted immediately (days, or hours) after a potentially stress-inducing major national event. The results from these surveys would be instantly available to researchers, who could change when and how the next question is asked. These are just a few possibilities of how a real-time sampling system might work.
Extending the System Capabilities
Once NESP participants have become comfortable with self-report data collection, they may be asked to opt-in to an additional sensor-data sharing program that sends summarized (and appropriately identity masked) sensor data to researchers. Data from new sensors embedded in phones (e.g. cameras, microphones, bar code scanners, etc.) or sensors specially-developed for the GEI that communicate wirelessly with the phones could be acquired and summarized, but also used to trigger other context-sensitive questioning. For example, nutrition researchers could send a prompt to subsets of hundreds or thousands of people at dinnertime, asking them all to take a photo of any food they are currently eating, which would be sent back to researchers to analyze. Stress researchers could ask people to record a sample of the noise level in their current context.
This is the vision motivating this project. Our primary aim is to use participatory design to develop these ideas, working with small panels of typical phone users and experts in health research.
What do you think? If you have a mobile phone, would you sign up for NESP? If you are a researcher, can you think of data that NESP might provide that would accelerate your research? Let us know.