Kolja Sam Pluemer

Intelligent Tutoring Systems: Entry Points

An overview of ITS

23.07.2024

This is an introductory post for anything looking to dive into Intelligent Tutoring Systems, computer-based learning applications enhanced by Artificial Intelligence, however that may look. It’s not comprehensive, and I’m not an expert — if you find any errors, please point them out to me.

  • First of all, the Wikipedia article provides a decent overview, especially regarding the concepts history [0].

  • To make sense of the research’s progression, it’s quite interesting to look at ITS from the perspective of AI hype cycles and winters — the topic is intertwined with the emergence of the term “AI” itself [1], was prominent in the AI surge of the 1980s [2] and is again going strong in the current era [3].

  • In my opinion, it is worth reading one or two old-school papers on the topic. The typewriter-set papers, often mainly considered with ruthlessly optimized hard-wiring of their logic and the (from my perspective) quite alien understanding of AI back then. [4] is full of those.

  • One of the most prominent ITS out there appears to be the Carnegie Learning’s Cognitive Tutor, which has about a dozen papers dedicated to it (for example [5]). Worth checking out.

  • ITS research often offers powerful mathematical tools to solve educational problems that you will not necessarily find in other research, like the Bayesian Petri net in [6] or the Adaptive Navigation Maps in [7].

  • A rather big problem of ITS is that implementations are often hard to access. Using them to learn or teach is difficult, but even finding a simple screenshot or demo is often not easy, as you will find when combing through the examples above. OSS implementations are even rarer [8].

  • Over time, you will certain recurring terms in ITS research (ITS term): Very prominent is the idea of models, such as domain model, student model, pedagogical model and adaption model. Another interesting idea are learning bugs.

  • As expected, ITS do not stand alone. You may also be interested in the following keywords — some describe subcategories, others are umbrella terms, others are merely similar: Cognitive tutor, CAI, Adaptive Instructional Systems, Programmed Instruction, Teaching Machine, computer-as-tutor/computer-as-tool. ACT-R and Spaced Repetition in general also seems like compatible ideas, although I’m not sure of the actual overlaps. If you’re looking for a more general term describing what both of those are trying to do, try instruction sequencing.

Terms

  • ITS: Computer-based learning supported by AI [9]
  • AI: Artificial Intelligence
  • OSS: Open Source Software
  • adaption model: decide which learning objects to display
  • Cognitive tutor: a special kind of ITS, see for example [10]
  • instruction sequencing: which exercise should be scheduled when