Presentation on theme: "Cutting Through the Crud How BCI and Recommender Systems Might Improve the Downhill Skiing Experience."— Presentation transcript:
Cutting Through the Crud How BCI and Recommender Systems Might Improve the Downhill Skiing Experience
Ski Equipment Selection Brands: Atomic, Dynastar, Fischer, K2, Rossignol, Volkl, Salomon, Tecnica, Head, Tyrolia, among many other large and boutique brands. Ski styles: carving, powder, all-mountain, twin tips, integrated, big mountain, racing, among others. Boot styles: rear entry, mid entry, front entry. Binding types: variations in heel and toe release – diagonal heel, forward heel, spheric heel, upward toe, twist toe, etc. Within these categories, there is enormous variety in design, construction, and composition to match specific user needs based on body type, technique, skill level, terrain, compatibility with other equipment, etc.
Information Gap Buying new equipment is a significant investment. Skiing is dangerous, so its important that it fits well. Most people make a purchase with very little information exchanged. Skier can only generally describe ability and desires and fitter has to interpret and take information at face value. No skier history to work from. Theres usually no opportunity to try skis, boots, bindings on snow beforehand. Even if s/he does, only can sample a small subset of whats available. Like cars, ski equipment loses its value quickly with use. If theres a mismatch between skier and equipment, it is likely to persist for a season or more.
Recommender System: Skier Profile A recommender system based around well-developed skier profiles could be an excellent way to manage this problem. Basic skier profile would be: –1) body type – height, weight, foot size, etc. –2) list of past equipment purchased, along with satisfaction level –3) record of mountains visited –gauge skiing frequency –4) lifts taken at those mountains –gauge ability –5) responses to questions that look for behavior in specific skiing scenarios –gauge ability, technique, skiing temperament, etc. A profile could be created online, at a ski shop, or at a ski resort. Shops and resorts would have access to these profiles to update data: a ticket purchase would register mountain visit, scanning of ticket would register lift taken, etc.
Recommender System: Demos, GPS, and BCI In order to recommend equipment through the system, the tester would have to demo it at a ski resort. In addition to demoing skis, boots and bindings would be made available for evaluation. Tester would have to wear a helmet equipped with EEG sensors, a GPS device, and sufficient memory to record data from both throughout the day. At the end of the day, the tester would be asked to answer various questions on a numerical scale, as well as add comments. Question responses would be weighted positively or negatively based on correlation between them and helmet feedback, giving the review more or less credibility. This data would be added to the user profile, informing skill level rating, reviewer value, etc.
Recommender System: Other Details In addition to user profiles, there would be extensive specs on all skis, boots and bindings available, so that associations between different equipment could be made by the system. Users would have the ability to rate the feedback provided by others after theyd made a purchase of the same or similar equipment, giving strength to the reviews provided by that individual. The structure of the user interface for the system would be akin to a social networking site, which the ability to add trip pictures and so forth, which would add incentive to contribute.
Some Challenges Getting the first users to contribute to the system, since theres initially no incentive beyond curiosity. Utilizing data in an effective way algorithmically so that feedback and recommendations are accurate and valuable. Getting BCI technology to the point where it can be used effectively in this way and not be overly intrusive or unnerving to users.