Wildlife Master Volunteers: A Multi-County Approach to Resolving Human-Wildlife Conflicts

The Wildlife Master (WM) Program in Colorado was modeled after the highly successful Master Gardener volunteer program. In 10 highly populated suburban counties with large rural areas surrounding the Denver Metro Area, Colorado State University (CSU) Cooperative Extension Natural Resources agents train, supervise and manage these volunteers in the identification, referral, and resolution of wildlife damage issues. High quality, research-based training is provided by university faculty and other professionals in public health, animal damage control, wildlife management and animal behavior. Inquiries are responded to mainly via telephone. Calls by concerned residents are forwarded to WMs who provide general information about human-wildlife conflicts and possible ways to resolve complaints. Each volunteer serves a minimum of 14 days on phone duty annually, calling in from a remote location to a voice mail system from which phone messages can be conveniently retrieved. Response time per call is generally less than 24 hours. During 2004, more than 2,000 phone calls, e-mail messages and walk-in requests for assistance were fielded by 100 cooperative extension WMs. Calls fielded by volunteers in one county increased five-fold during the past five years, from 100 calls to over 500 calls annually. Valued at the rate of approximately $18.00 per volunteer hour, the leveraged value of each WM was about $450 in 2005, based on 25 hours of service and training. The estimated value of the program to Colorado in 2004 was over $45,000 of in-kind service, or about one full-time equivalent faculty member. This paper describes components of Colorado’s WM Program, with guides to the set-up of similar programs in other states.

Bandit-Based Algorithms for Budgeted Learning

We explore the problem of budgeted machine learning, in which the learning algorithm has free access to the training examples’ labels but has to pay for each attribute that is specified. This learning model is appropriate in many areas, including medical applications. We present new algorithms for choosing which attributes to purchase of which examples in the budgeted learning model based on algorithms for the multi-armed bandit problem. All of our approaches outperformed the current state of the art. Furthermore, we present a new means for selecting an example to purchase after the attribute is selected, instead of selecting an example uniformly at random, which is typically done. Our new example selection method improved performance of all the algorithms we tested, both ours and those in the literature.