Ask a Librarian Bot
On the way to work this morning I listened to one of my favorite podcasts, Radiolab. The most recent episode, “Talking to Machines,” describes various robots that attempt to behave like humans. Rolla Carpenter tells the hosts about his creation, Cleverbot, a chatbot that has been learning from conversations with humans over the last 10 years. Ask Cleverbot a question and it checks a database of millions of past conversations with humans and pulls the most relevant response, something Jad Abumrad of Radiolab eerily compared to “talking to a ghost.”
This got me thinking about Ask a Librarian chat. It is common for academic libraries to offer some form of chat reference. It’s considered a convenient and modern service for hip young college students who prefer online communication over in-person, something I’ve come to value as well (since I’ve discovered Seamless Web I will never call a restaurant for delivery again!). Could we provide a database to a librarian bot that would be sufficient to answer the basic reference questions of our patrons? Would the bot be intelligent enough to know when a question was too complex, and refer the patron to set up a consultation?
Right now we contract our reference chat service with Tutor.com. In 2010, these trained librarians with access to TC and CU databases completed almost 500 online reference sessions. No offense to the Tutor.com employees, but could they be replaced by a bot? In compiling reports on Services at Gottesman, I look at the kinds of questions that are most frequent for our Tutor.com librarians. The majority of them are:
- Questions about physical access to the library or online access to resources.
- Requests for help, such as advice on search terms and databases to use.
- Circulation-related questions.
- E-Reserves-related questions.
- Where can I download/how do I use Endnote?
- Do you have this article/journal/book? If not, how can I get it?
- Questions about room reservations
In my brief experience I learned that chatbots are bots of few words. They don’t elaborate, probably because if they do, they will be revealed as artificial. Humans don’t tend to have conversations in exclusively complete sentences, but that is what the chatbots need to provide relevant answers. If you ask a follow-up question, make sure you put it in context. The most comical example for me came after only a few seconds of talking to Cleverbot:
When I tried to clarify Cleverbot got a bit of an attitude with me, saying “You really like to change the topic around do you?” I never did find out if it preferred Ibuprofen or Acetaminophen. Chomsky was more fun. It actually provided more information than necessary, unlike these other Jordan Catalano-bots that stick to one word answers. When I asked Chomsky if it goes to the library, it said it goes there to chat with humans about philosophy. “To ridicule philosophy is truly to philosophize,” it said. No doubt some TC students could turn this into intriguing robot discourse. I didn’t expect Chomsky to be the winning conversationalist since the database it’s accessing is Wikipedia, not 10 years of human conversation like Cleverbot.
At Gottesman we already have a number of databases to pull from that could help our librarian bot. The Knowledge Database is full of information and FAQs. Wolfram Alpha is a computational search engine that is particularly good at answering math and science questions. And, while I don’t expect any TC patrons to come to the library to ask something like “How to get paint off my trousers???,” adding Yahoo Answers, Ask Metafilter and the like would provide the librarian bot with a possibly overwhelming overview of the widest range of issues on the “hive mind.”
browser to pages relevant to your queries. Personally, though, I have seen these customer service bots before and I tend to ignore them because I feel, maybe wrongly, that they will not be able to help me as a human could. Perhaps the key is to implement our librarian bot inconspicuously and see if people can figure it out? After all, as Robert Epstein told Radiolab, he was fooled twice by chatbots on dating websites, convinced for a time that the bots were real girls. The idea that robots can take over librarians’ jobs is not new–just watch the Desk Set. I believe, as Katharine Hepburn discovered in the 1957 film, that humans will have a place in libraries for years yet; but wouldn’t it be nice to have robot helpers?