Mobile‐Enhanced Field Research Increasing plant identification accuracy and efficiency Autumn 2013 David Gagnon, Seth McGee, Breanne Litts, John Martin, Justin Moeller, Nick Heindl, Phil Dougherty
Abstract In this report, we discuss a pilot study of a plant identification field research activity in an undergraduate “Biocore: Ecology, Genetics, and Evolution” laboratory course at the University of Wisconsin‐Madison. This study investigates the efficiency and accuracy of using mobile devices for field research data collection activities. We draw from theories of distributed cognition (Cole & Engeström 1993; Hutchins, 1995) and situated learning (Brown, Collins & Duguid, 1989; Lave & Wenger, 1991) to explore how novices use mobile devices in the field to collect data. We conducted a design experiment (Brown, 1995), and we offer our findings and their implications for the development of a generalized field research platform. Keywords: Biology; Mobile; Citizen Science; Fieldwork; Plant Identification
http://mobile.wisc.edu/teaching‐and‐learning/
1
Purpose We use the following case to achieve these goals: 1. Compare mobile technology with traditional methods for field research activities primarily focusing on answering these two questions: a. Can using mobile devices make field research data collection more efficient? b. Can using mobile devices make field research data collection more accurate? 2. Inform the development of a generalized field research platform. Overall, the purpose of this pilot study is to demonstrate some of the affordances and constraints of using mobile technology to complete field research activities. As we continue to experiment with the design and implementation of field research activities, this study also informs our effort to build a more robust mobile field research platform.
Context In Autumn 2013, we piloted a mobile‐enhanced field research activity with Seth McGee, co‐instructor of the Biocore 301/302: Ecology, Genetics, and Evolution laboratory courses at University of Wisconsin‐Madison. As part of these courses, students are required to participate in a variety of field activities which allow them to learn about the ecology of native Wisconsin ecosystems. We used three laboratory sections (10 students in each) that focused on plant identification to conduct our study.
Theoretical Frame We take up a situative perspective to understand how and what students make meaning (Brown, Collins & Duguid, 1989; Lave & Wenger, 1991). Thus, we define learning as something that happens in context and community. More specifically, we believe that learning happens through participation in authentic practices and discourses of a given community (New London Group, 1996; Wenger, 1998). In this study, we situate learners in an actual biocore prairie using expert discourse to identify real plants. Learning, in this view, is a process of increasingly becoming central to the biology community through their practices and discourse to authentically engage in the discipline. Additionally, we also define learning through a lens of distributed cognition (Cole & Engeström 1993; Hutchins, 1995); in particular, learning broadly extends from the learner into the world through tools (Norman, 1993). Put differently, we believe that learning is a knowledge‐sharing process between learner and tools and/or artifacts. With this study, we are designing a new tool (i.e. a mobile app), and effectively test whether this app is more efficient and accurate as a distributed learning artifact than the traditional field guide.
http://mobile.wisc.edu/teaching‐and‐learning/
2
Tool While the application developed for this study was inspired by the traditional dichotomous key approach used in many field guides, it also draws from the unique affordances of mobile app design and user experience conventions. The main screen of the app presents a number of plant anatomy components , made up of a small icon and a title. These include items such as “Leaf Shape” and “Flower Color.” Upon touching one of these categories, a selection screen appears that contains a collection of fixed options such as “Elliptic” and “Oval” with an associated illustrations. For open‐ended or numerical components, such as flower width, a text entry field is presented instead. Immediately after the user makes a selection or enters information, the selection screen slides offscreen and the main anatomy screen screen is displayed again, now indicating the user’s selection for one of the components. A prompt at the top of the screen concurrently updates with the number of possible species that match the user’s selection. Each time a user adds a new selection, the number is reduced. Touching the “Identify” button below the “Possible Matches” label displays a list of all the species in the app, sorted by the likelihood of a match. Each species has a details page, featuring descriptive information and a series of relevant images.
http://mobile.wisc.edu/teaching‐and‐learning/
3
Data We take up a Design‐Based Research approach for our overarching exploration of field research activities, and, as part of that, we conducted a more formal design experiment (Brown, 1992) in this study. Specifically, we compared across three conditions of learner‐tool relationship: mobile device only, mobile device and book, and book only. Students volunteered to participate in one of the three field research sessions with no knowledge of which condition they would be participating in — or what the study was about. In each condition, five pairs of students1 completed the same task of identifying ten plants in the UW‐Madison BioCore Prairie. Further, to determine how mobile technology affects the practice of field research activities, we were interested in comparing differences of efficiency and accuracy across conditions. To capture both of these variables, students were required to fill out a worksheet (see Appendix 1) naming each plant and reporting the time they began, and the time they finished, identifying the plant. Additionally, to evaluate how mobile technology impacts students’ discourse about plants, we compared students’ self‐generated content‐related vocabulary across conditions. Overall, we collected a wide range of data for explorative and informative purposes: ‐ Self‐reported start and stop time of identifying each plant (efficiency) ‐ Expert evaluations of students’ identifications (accuracy) ‐ Post‐survey capturing their experience and knowledge of plant identification (see Appendix 2) ‐ Post‐field activity content knowledge task: students were asked to, “Write or draw any words or things you can think of related to plant identification.” ‐ Observation and focus group data to determine how students engage with mobile devices in the field. In this report, we discuss findings from the field activity worksheet and post‐field activity content knowledge task to highlight some of the affordances and constraints of using mobile as a field research tool. Additionally, we draw from post‐survey, observation, and focus group data to inform the development of future field research tools.
Mobile technology versus book technology Across quantitative items (efficiency, accuracy, and content knowledge), we used a Kruskal‐Wallis test to determine differences between the three conditions. Where appropriate, we followed up with a Tukey’s HSD post‐hoc test to specifically identify where differences lie. Efficiency 1
The “both” condition had six pairs of students, but one pair was not enrolled in the course, so they were dropped from analyses where possible.
http://mobile.wisc.edu/teaching‐and‐learning/
4
Using mobile devices in field research activities can help users filter information much quicker than traditional book‐based methods. In this study, we sought out to determine whether this suspicion held in practice. As students completed the plant identification activity, they recorded their start and stop time for each plant. Pairs using an iPad took an average of 43.4 minutes to complete the task; pairs using both an iPad and a book took an average of 57.8 minutes to complete the task; and pairs using only a book took an average of 78.8 minutes to complete the task. Our statistical analyses revealed that there is a significant difference [p=.000] between the time it took for iPad‐only and book‐only pairs to complete the task, and a significant difference [p=.010] between the iPad‐and‐book and book‐only pairs. Thus, providing novices mobile devices to complete field research activities appears to increase their efficiency at accomplishing identification tasks.
Accuracy In addition to efficiency, it is equally important to determine whether mobile devices improve accuracy of field research. Accordingly, an expert botanist evaluated whether students correctly identified each of the ten plants. Pairs in the iPad‐only condition averaged 72% accuracy; pairs in the iPad‐and‐book condition averaged 66% accuracy; and pairs in the book‐only condition averaged 20% accuracy. Both the iPad‐only and iPad‐and‐book conditions achieved significantly higher accuracy in their plant identifications than the book‐only condition, p=.007 and p=.016, respectively.
http://mobile.wisc.edu/teaching‐and‐learning/
5
Content knowledge At the end of the plant identification activity, we asked students to write or draw terms or ideas they could think of in relation to plant identification. Interestingly, students in the iPad‐only and book‐only averaged a similar number of words or drawings, 5.1 and 7.2, respectively. Yet, the students using both the iPad and the book averaged 13.8 words or drawings—significantly higher than both other conditions, p=.000 for both. This means that, perhaps, having two resources available significantly improves learning, as defined as picking up and using expert discourse of plant identification . While there are a number of possible interpretations of this finding, one clear difference is the one‐to‐one ratio of students‐to‐resource in the iPad‐and‐book condition. Simply, having two resources available to each pair may have a greater impact on their ability to build discourse. To explore this difference further, we think a next step might be to compare groups where every student has an iPad and/or book, rather than pairs of students. The qualitative data we collected throughout this study suggests that pairs who had both resources more richly discussed and negotiated their identification of plants (e.g. students were constantly checking and confirming with each other).
Informing a field research platform Our goal in this pilot study was twofold: to investigate how mobile devices support novices’ development and to determine the affordances and constraints of the current plant identification tool with an eye toward expanding to a more general field research platform. Scaffolding across varying skill levels We originally designed the plant identification app to help novice’s complete real identification work in the field; however, we learned through this pilot study that we could scaffold this process to suit a wider range of experience and expertise. In our current app, users select features of the plant
http://mobile.wisc.edu/teaching‐and‐learning/
6
until they’ve narrowed down their options enough to successfully identify the plant. This onramp worked well for extreme novices, but many students requested a search function for “backwards identification”, as they were able to successfully recognize some plants on sight, but wanted to use the app to double‐check their suspicion. Dynamic and crowdsourced information The benefits of using a digital versus analog tool in the field is that information can easily be added, changed, and rearranged. For instance, one of the most commonly reported challenges with using books for plant identification is that they are organized by flower type, and plants only flower during certain seasons. Students in the iPad‐and‐book condition strategically drew from both resources using the app to identify the plant and the book to confirm their identification, using it’s existing images and descriptions. From an accuracy perspective, a number of users reported that they highly depended on the photos of individual species for their final assessment in the field. They wanted as many photos available as possible from many angles, focal points and over the various seasons. They also wanted to see “look alike” photos that would highlight differences between species that are very similar. Hence, for them, the app wasn’t just about identifying plants, they also wanted to use it to learn more about plants in context of the identification activity. While it is certainly possible to pre‐populate an app with more media, another option for future development would be to create a system for user collected images to be integrated into the canonical identification media after it had been reviewed. It is also possible to integrate a look‐alike field into the database. Students seemed to be hungry for this information. Our goal for future iterations of this project is to build a platform that efficiently supports the identification process and offers relevant information about each plant. User experience As with any pilot, we were also able to identify a number of bug fixes and user interface improvements for the app. The depth of use data we collected on the functionality of the app itself has vastly enhanced our understanding of how students use mobile field research tools in the field. The design of app’s sorting algorithm as well as the user interface could be improved to speed up the interaction. The computations that are performed to rank and sort the possible species need improvement as they can take over 2 seconds to perform. Users reported that apps should have no noticeable delays. A few users also suggested that that the main screen with the different anatomy components could be combined with the screen that allows the user to specify a particular option, saving time in the transitions. Users also suggested that the app could pre sort the anatomy options based on the season in which the observation is being made. For example, show flower shape and color options if many plants are in the blooming season, which would quickly narrow the likely species.
http://mobile.wisc.edu/teaching‐and‐learning/
7
A feature that was dropped in the last moment, for the sake of testing the identification mechanic, was networked observation logging. Since the app already has a camera, GPS, keyboard, etc. it is a perfect opportunity for recording the pure observational data at a location, as well as the user’s interpretation of that data to determine a species. For grasses and flowers, the density, not the existence of a given species in a region is the measurement used in the discipline.
Takeaways In this pilot study, we aimed to evaluate the efficiency and accuracy of using mobile technology for field research activities. Not only did we find solid evidence that both variables are significantly improved through the use of our iPad app, but we also found evidence suggesting a one‐to‐one ratio structure of implementing field research activities. Put differently, students were able to more efficiency and more accurately identify plants using mobile devices, and perhaps giving a mobile device to every student — even if they’re working in pairs — could significantly impact the content knowledge they develop. With this, we are designing field research activities that immerse learners in both the practice and discourse of actual scientists in a given field. Thus, we gathered a range of feedback and evaluative data of the field research application in order to further the development of a generalized field research platform.
References Brown, A. L. (1992). Design experiments: Theoretical and methodological challenges in creating complex interventions in classroom settings. The journal of the learning sciences, 2(2), 141178. Brown, J. S., Collins, A., & Duguid, P. (1989). Situated cognition and the culture of learning. Educational researcher, 18(1), 3242. Cole, M., & Engeström, Y. (1993). A culturalhistorical approach to distributed cognition. Distributed cognitions: Psychological and educational considerations, 146. Hutchins, E. (1995). Cognition in the Wild (Vol. 262082314). Cambridge, MA: MIT press. Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheral participation. Cambridge university press. Lave, J. (1991). Situated Learning in Communities of Practice. In L. Resnick, J. Levine, & S. Teasley (Eds.), Perspectives on socially shared cognition (pp. 6382). Washington, DC. American Psychological Association. New London Group. (1996). A pedagogy of multiliteracies: Designing social futures. (C. Cazden, B. Cope, N. Fairclough, J. P. Gee, M. Kalantzis, G. Kress, A. Luke, et al., Eds.)Harvard Educational Review, 66(1), 60–92. Wenger, E. (1998). Communities of practice: Learning, meaning, and identity. Cambridge university press.
http://mobile.wisc.edu/teaching‐and‐learning/
8
Appendix 1: Biocore Worksheet
http://mobile.wisc.edu/teaching‐and‐learning/
9
Appendix 2: Biocore survey
http://mobile.wisc.edu/teaching‐and‐learning/
10
