|
Your Caffeinated Professor!
This professor (there in the lower right-hand corner) never sleeps, and never gets tired of answering questions. That’s the good part. The bad part: because he never sleeps (and because he's AI) his answers sometimes go berserk…
You're the human, so it’s on you to figure out when the AI is helping, and when it's babbling.
Finally, you are anonymous here. Using the model doesn't require any personally identifying information. However, we do collect and use questions and responses for model improvement.
This AI has been trained to answer questions in exactly the same way that Professor Brusseau would, within the context of the Philosophy 121 course, Ethics in the Workplace. The tool has been trained with the course textbook, and with transcriptions of recorded lectures. For that reason, the machine works well sometimes. But, when questions are posed in ways that do not cohere well with the training data, results can be divergent.
There's no way to predict beforehand which questions will work well, and which ones will short-circuit. The machine is imperfect, and students must weigh the AI output against the course’s lectures and assigned readings. In case of conflict, the human generated content and material always takes precedence over the automatic content.
The tool cannot be used reliably for procedural information (due dates, exam dates, etc.).
The purpose of this model is to increase accessibility for student questions in two ways. First, questions may be posed 24/7. Second, the questions are anonymous. For a variety of reasons, students may hesitate to present themselves and ask questions of the human professor. They may not be sure what they want to ask, and need to begin questioning generally as a way of identifying their own specific uncertainties. Other times, students may feel self-conscious about posing questions because they fear they are being judged by the professor. Or, they may not want to reveal their own misunderstandings of the course. In cases like these, students may benefit from anonymous, automated responses.
The tool does not replace the professor, but it does seek to increase the course’s accessibility, effectiveness, and efficiency.
Model Card: The Caffeineted Professor
What is it?
The Caffeineted Professor is an AI conversational agent that has been trained to answer questions in the same way that Professor Brusseau would, within the context of his Philosophy 121 course, Ethics in the Workplace.
The agent is available online 24/7. (It never sleeps)
Why did we build it?
The model’s purpose is to increase accessibility for student questions in two ways.
1. Questions may be posed 24/7.
2. Questions are anonymous. For a variety of reasons, students may hesitate to present themselves and ask questions of the human professor. They may not be sure what they want to ask, and need to begin questioning generally as a way of identifying their own specific uncertainties. Other times, students may feel self-conscious about posing questions because they fear they are being judged by the professor. Or, they may not want to reveal their own misunderstandings of the course. In cases like these, students may benefit from anonymous, automated responses.
The tool does not replace the professor, but it does seek to increase the course’s accessibility, effectiveness, and efficiency.
Who built it?
James Brusseau (PhD, Philosophy) is a professor in the Philosophy Department at Pace University in New York City during the fall and winter, and he is a visiting professor in the Computer Science Department at the University of Trento in Italy during the spring and summer. His academic research explores the human experience of artificial intelligence in the areas of personal identity, authenticity, and freedom.
Weijia Huang, a Boston University graduate, has been programming for over 10 years, specializing in AL and ML. He is passionate about innovating education with technology.
How did we build it?
The conversational agent was trained to mimic Professor Brusseau in the Philosophy 121 course, Ethics in the Workplace, with the course textbook authored by Brusseau, and with transcriptions of Brusseau’s recorded lectures. Subsequently, computer generated questions and answers were annotated by Brusseau, and then algorithmically augmented by a 10 factor. The foundational model for 1.0 was the OpenAI Generative Pretrained Transformer.
The model for this version 2.0 is Deepseek. The model-building and fine tuning is iterative. Student questions and the tool’s responses will be reviewed to supplement the training data.
|
