The Cornell Lab of Ornithology (CLO) is creating a new type of interactive, question-driven, online bird-identification tool called "Merlin," along with associated games, social networking tools, and other media. Unlike existing bird-identification guides, which are based on traditional taxonomic keys written by scientists, Merlin uses machine learning algorithms and crowd-sourced data (information provided by thousands of people) to identify birds and improve Merlin's performance with each interaction. The tool will help more than twelve million people a year identify birds and participate in a collective effort to help others. The Crowd ID project will make it easier for people to discover the names of birds, learn observation and identification skills, find more information, and appreciate Earth's biodiversity. The summative evaluation plan is measuring increases in participants' knowledge, engagement, and skills, as well as changes in behavior. Impacts on participants will be compared to a control group of users not using Merlin.
Crowd ID tools will be integrated into the CLO's citizen science and education projects, which reach more than 200,000 participants, schoolchildren, and educators across the nation. Merlin will be broadly adapted for use on other websites, social networking platforms, exhibits, mobile devices, curricula, and electronic field guides. Once developed, Merlin can be modified to identify plants, rocks, and other animals. Merlin will promote growth of citizen science projects which depend on the ability of participants to identify a wide range of organisms.
Project Outcomes With the majority of American adults now using smartphones, the opportunity to use mobile tools to engage more people in "anywhere, anytime" STEM learning has never been greater. The Merlin Bird ID app takes a novel approach to helping people identify and learn about 400 of the most commonly encountered bird species in North America, using their mobile devices. Unlike field guides or online keyword searches, the Merlin app provides a guided experience, asking the user five questions about the bird they saw. It taps into 70 million observations from the eBird citizen-science project to determine the most likely species customized to the user’s location and time of year. Recognizing that different people describe birds in different ways, it also uses 3 million crowdsourced descriptors of birds, contributed by the public, to find the best matches. Machine learning techniques were developed to improve Merlin’s performance through time, advancing research on human-computer interactions and the use of variable crowdsourced observations to return accurate results. To assess Merlin’s effectiveness, a summative evaluation was conducted comparing beginner and novice bird watchers who used the web and/or app version of Merlin compared with a control group. Participants who used Merlin perceived greater gains in bird identification and observation skills. Those who had access to both the web and app versions perceived the strongest gain. Merlin users who had access to the app were more likely than the web-only group to state they believed: Merlin increased their bird observation skills. Merlin increased their bird identification skills. Merlin was effective in helping them learn more about birds. They were more likely to watch birds again because of Merlin. They believe Merlin is a useful tool for future bird watching. Participants who had access to the app and web version clearly prefer the app, primarily because of portability and utility. Participants frequently commented on their enjoyment of the app, 88% said it was easy or very easy to use, and 90% said they would use it again. Parents also reported using it to identify birds with children. In interviews, participants commented that Merlin was quicker, easier, and more immediate to use than field guides or Internet searches. There is supporting evidence that Merlin improved observational skills: individuals who perceived an increase in their bird watching skills were clear that Merlin trained them to focus on specific characteristics, often different than those they had previously noted. Those who had used Merlin showed a marked increase in using size as a clue, and slight increases in citing location and date--three of the five characteristics that Merlin emphasizes. Merlin users were also much more likely than the control group to explore more information on the Cornell Lab of Ornithology's All About Birds website, indicating engagement and the desire to learn more. Even though Merlin app users reported perceived gains in skills and showed evidence of which skills had increased, in tests of skill-based identifications using an online bird identification quiz, results were mixed. It is possible that asking participants to watch birds for as little as three times (at least once a month for three months) was not enough to see demonstrable change in skills using the 13-question quiz. There was some smaller amount of positive change in Merlin users, indicating that with perhaps a longer study, greater impact might be demonstrated. Nonetheless, users reported great pleasure in using the app, and it is an effective tool for beginners and novices to identify birds. The Merlin app is available for free to the public and was downloaded by more than 325,000 people and used in more than 4.3 million sessions in the first year. Users have confirmed more than 1 million bird identifications through use of the app. In 2014, Merlin won the Gold Award for best mobile app from the Council for Advancement and Support of Education (CASE) and was an Appy Award finalist for best reference app. Merlin is helping to engage the public in learning about wildlife, which is especially important at a time when the Earth is facing a biodiversity crisis; 30% of species assessed with ICUN Red List criteria are threatened with extinction. In addition to helping people learn about birds, Merlin serves as a model that can be adapted to help people identify other wildlife species, such as butterflies, frogs or fungi. Merlin has the potential to support public participation in science because accurate species identification is a basic requirement of most biodiversity studies. Many users requested the ability to keep track of the species they identified, suggesting the power of using mobile technologies to bridge lifelong learners from curiosity to observation and documentation, essential steps in the process of science.
