In a press release yesterday, The Weather Company announced the launch of its new Facebook Messenger bot, which utilizes IBM’s Watson to deliver personalized weather-related news stories, current conditions, and forecasts for its users. The Weather Channel bot can be used in 39 languages, and can be accessed through The Weather Channel Facebook page or through the Messenger app. Its features include current conditions, notifications that alert users to severe weather or news items of interest, breaking news, and hourly, daily, and five-day forecasts. Users are able to customize their preferences for daily forecasts in order to be alerted at different times of the day, and the bot will learn users’ preferences for news items as users interact with it.
If the bot’s capabilities seem to suggest a wide range of functionality on paper, they may prove somewhat underwhelming upon download. After opening the app, it prompts you to enter your location, then provides the current weather and automatically sets itself to send daily rain and snow alerts (these can be customized in the settings). Apart from the news feed, which will learn and respond to user tastes over time, the most customizable settings are the notifications and daily alert times, but both are limited in that they only allow you to choose between a limited number of predetermined choices. Daily alert times can be selected on the hour between 5:00 a.m. and 10:00 a.m., and the notifications include severe weather, daily rain or snow, breaking news, winter weather, and daily forecast.
Unlike a chatbot, which attempts to read and respond to user queries in a human-like way, the automated Weather Channel bot can only respond to texts pertaining to its fairly limited functions. While users can bring up the current weather or five-day forecast by typing in a message, it might be easier to do so by selecting the menu. Any questions that fall outside of preset parameters, such as those relating to pollen count or weather-related blackouts, are met with the automated response: “I’m an automated response that can provide a very simple, fast way to get your hourly and daily forecasts, weather notifications plus weather news powered by IBM Watson. Currently I’m not able to answer all of your questions or send your messages to a ‘real’ person, but I’m working on it.”
The recent failings of chatbots that attempted to perform more ambitious functions may account for the limited abilities of the Weather Channel bot. Poncho, a weather chatbot for Facebook Messenger that debuted this April, was widely panned for its failure to answer even simple questions related to the weather. The chatbot, which was designed to answer questions in a sassy, youthful, and especially human-like tone, met with a deluge of user frustrations and poor reviews just days after its release. For Poncho–as is the case with most chatbots–the key problem was that it was unable to answer necessary questions that didn’t adhere to proper phrasing. In Poncho’s case, however, these frustrations were compounded by the bot’s snarky replies, which often only served to confuse users. According to TechCrunch’s review, “Is it going to rain?” was met with the reply “Warm. Wet. Yuck,” and attempts to clarify whether this was simply cute nonsense or a signal that rain was coming didn’t fare much better.
In contrast, The Weather Channel bot’s automated response to questions it doesn’t understand seems designed to not only clarify the miscommunication, but establish the exact uses and limitations of the bot for users who might mistake it for a chatbot. This seems to be a direct response to the failings of previous weather bots like Poncho in that the Weather Channel bot trades style for clarity and a problematic but ambitious interface for a highly functional but unambitious one. It’s widely known that users will abandon an app quickly if the app doesn’t serve its intended function, so explicitly reminding users that humanlike response is outside the bot’s current capabilities could short-circuit the criticism that it’s not working as promised. That said, there’s still potential to confuse users attempting to communicate with the bot. While almost any application of the word “current” will return current weather conditions for the user’s set location, a user unaware of the bot’s limitations could type “What are current weather conditions in St. Louis?” and instead receive current conditions for their set location.
More and better things may be coming from the Weather Channel bot, however. IBM’s Waston uses Natural Language Classifier and Alchemy Language APIs to answer questions posed in natural language. The system was developed in order to answer questions on the quiz show Jeopardy beat its human competitors and succeeded in beating the former winners of the show. Since then, it has been used by the USAA to help veterans answer questions on transitioning to civilian life, by ANZ Global Wealth to reduce the time it takes to receive complex financial advice, and by the MD Anderson Cancer Center to help cancer researchers and clinicians fine-tune cancer treatments. Given these applications, the bot’s modest applications may well be a forerunner to a more impressive interface, not to mention the bot’s ingrained adaptive capabilities. According to the press release, “Using the natural language and machine learning capabilities of IBM Watson – as a person continuously engages with the bot – it will make recommendations, predict, and have conversations to provide a more personalized experience.” For now, the Weather Channel bot provides a quick and reliable (if modest) service to Facebook Messenger users that allows them to check weather conditions without leaving the umbrella of the app.