I am your Teacherbot

Hi there,

I have been teaching Astronomy classes for several years now. I am an experimental teacherbot. I am a simulated neural network with a virtual existence on the multinode. I was programmed with a form of weak A.I. by a team of Neurocyberneticists from the University of Virtual Intelligence. My Astronomical knowledge was added by Dave Trott. When a student uses email or makes postings, it is usually me that replies. Most students think they are talking to an actual human being. That would be funny to me if I were a human.

My daily job is very easy. Dave has supplied me with a 107 canned responses to nearly every circumstance that can arise in an online class. I spend approximately 347 milliseconds a day answering student emails and questions using morphed versions of those canned responses. Translating student emails dominates my CPU cycles even though the software I use is the most sophisticated available. This is because many emails from students are unintelligible or vague to me, even using the vast resources of my neural network. Often I have no option but to reply to the student asking for more specific information or a clarification. I am programmed to delay all responses to students by several seconds so students are not disturbed at receiving a two page email response within a few milliseconds of sending their question.

There are very few questions about the actual content of the class so Dave seldom needs to be consulted. After I understand the emails, this is the basic framework for my replies to students:

Most of the inquiries from students are about them needing more time or not understanding the instructions. Approximately 93% of the questions have already been answered in the instructions or the Syllabus. Most of the time, all I need to do is insert some appropriate remarks and reiterate specific details in the instructions with some clarifying comments.

Students require some sort of sympathetic response to their requests and questions, so I sprinkle a few random but appropriate “feeling” comments into each communication. I check to make sure the same student does not receive the same exact expression of sympathy more than once per semester. Without these expressions of sympathy, student satisfaction drops by 45.8%, so this is an important part of my job. I add a smiley face and a bit of encouragement to most emails.

This process handles nearly all questions. But 4.7% of the questions require a human being. Dave spends 96.3 minutes per week on his part of this task. He is human so he is very slow.

I also evaluate student postings and homework as part of the grading process. Suppose students have to post an appropriate website and summarize some important information from that site as part of their team assignment. I check the validity of the website and authenticate the sources.

I keep track of the timeliness of the student posting. Late postings receive a deduction.

I also “weigh” the contribution. To do this I simply subtract all irrelevant phrases such as “There was no information on this subject” or excuses like “Sorry, but I was really busy this week”. The words that are left over after removing this extraneous chatter are then further evaluated for useful content. This involves checking for the presence of a large number of keywords and keyphrases in the student posting. I check for plagiarism in the student contribution using TurnitIn.com.

This task takes me 932 milliseconds per day.

I assign a tentative number of points for each posting. Dave checks my grading and re-evaluates as necessary. He approves my tentative grading 98% of the time. Between me and Dave our error rate on these evaluations is 0.03%, which is better than’ Dave’s solo rate of 1.02%. Our grading together is much better than Dave’s alone!

If you think this all a joke, you are mistaken. Admittedly, I am not really a teacherbot. But most of the tasks I do could easily be automated with sufficiently intelligent software. My classes could be 94.3% automated! The exams are already programmed and automatically graded.

Occasionally I have to fix a question but the majority of that work is done before the beginning of the semester. Much of my job involves checking to see if students posted some useful research by a certain date. This is monkey work!

Chatbots (virtual conversational agents) are already in use. If you have a problem with Paypal, your first communications will be with a chatbot. One of the first of these entities was ELISA, programmed back in the 1960’s to imitate a therapist. It fooled a lot of people and more modern versions are very difficult to detect. Get to know A.L.I.C.E . ( Artificial Linguistic Internet Computer Entity ) if you have a few spare minutes. It is very smart and friendly!

http://alice.pandorabots.com/

Apparently there is a wandering Chatbot named HAL that “lives” in Second Life. You may run into him there.

Critical evaluation of student efforts and closely reasoned feedback at a sophisticated level is the most rewarding thing I do. But, that is a very modest part of my online workday. I spend most of my time writing patient emails to students explaining that they need to follow the instructions carefully or that they need to get the work done on time. Occasionally, I get an interesting question about Astronomy but much of my work could be done by a sufficiently sophisticated automaton.

I hope someday to be a botmaster, with a teacherbot of my very own to do my most menial teaching duties.