Artificial intelligence and education

Definition

Artificial intelligence and education refers to a research community that is interested in the intersection of artificial intelligence research, learning and education.

Typical sub-fields of study are intelligent tutoring systems, intelligent learning environments, adaptive hypertext systems, some computer-supported collaborative learning systems, etc.

Purpose and early History

This section is based on Dillenbourg (1994:1-2) with slight cuts and changes by DSchneider.

Different motivations led scientists to apply artificial intelligence (AI) techniques to educational software and training software. Courseware developers were seeking more powerful techniques to build educational systems. On the other hand, researchers in computer science and in cognitive psychology found an opportunity to develop and test new techniques or new theoretical models. This second line has probably been the most influential during the eighties. It led to major scientific contributions. For instance, designers transformed the expert system design to develop systems which fulfil the educational functions (explanation, diagnosis, ...) expected in a training software. This work contributed to the elicitation of strategic levels in expertise (Clancey, 1987) and, later, to the emergence of second generation expert systems (Steels, 1990). In others words, research on educational applications helped to develop the methodology for analyzing expertise (knowledge engineering). Similar contributions have been produced in cognitive psychology. The work on learner modelling (trying to infer what the learner knows or misunderstands) has been central to the formalisation and evaluation of cognitive models (Anderson et al, 1989).

(1) The major contribution of AI to educational technology is the possibility to model expertise, i.e. that the system is able to solve the problems that the learner has to solve. The system is knowledgeable in the domain to be taught. The interest of AI techniques is less their ability to produce a correct solution than the way that this solution is constructed. For instance, some complex AI systems have been design to model the resolution of simple subtraction such as '234-98', while any computer language can produce the correct solution (Burton & Brown, 1982).

(2) Modelled expertise enables the system to conduct interactions that could be not conducted if the system worked with pre-stored solutions. Although artificial intelligence was originally intended to reproduce human intelligence, educational use of AI techniques does not require that these techniques are the prefect image of human reasoning. More modestly, it requires that AI techniques support expert-learner interactions during problem solving. Some degree of similitude may be necessary if we want the expert to talk about its expertise in a way that can be understood by the learner. For instance, neural network techniques are considered a better model of human reasoning than rule-based expert systems but they could not communicate with the learner about the knowledge encompassed in each of its nodes. From a courseware perspective, the quality of AI techniques is not their degree of psychological fidelity but the extent to which they support interactions which are interesting from a pedagogical viewpoint.

(3) The types of interactions supported by AI techniques are important for some learning objectives. These interactions are especially relevant when the goal is to acquire complex problem solving skills. Other learning objectives can be pursued with simpler interactions techniques, like multiple-choice questions. Since the development of an AI-based software is more costly that standard courseware (especially, those designed with advanced authoring tools), these techniques should be used only when they are really required.

The original goal of AI was to develop techniques which simulate human intelligence, i.e. which simulate the reasoning process itself or, more modestly, the outcome of this reasoning process . Now, with some distance with respect to this early days of AI, we can say that the role of AI techniques in courseware is to not to simulate human intelligence per se. The techniques are used to support interactions with the learner. Modelling expertise enables the system to 'enter' into the problem with the learner, discuss intermediate steps, explain its decisions, and reasons on the learner's knowledge (diagnosis). The focus has moved from reasoning AS the learner to reasoning WITH the learner. This evolution is not in contradiction with studies of human development which tend to consider intelligence not as the result of static knowledge structures, but as a capacity to interact with the our social and physical environment.

Please read the original article for an explanation of these three points, especially the link between the model of expertise and the types of interactions. Otherwise, refer to sub-fields of study introduced in the definition section.

See also cognitive tools and computer-supported collaborative learning (a field that attracted a lot of former AI&Ed researchers). A good journal (needs subscription) is the Journal of Learning Sciences

A view in 2016

Pierre Dillenbourg in The Evolution of Research on Digital Education asks how AI&Ed compares to 25 years ago. The paper identifies six trends and two "meta-trends":

1. More physical
2. Less semantic
3. More social
4. Less Design
5. More Open
6. More Teachers

The meta-trends are:

1. evolution of learning technologies has been mostly driven by the evolution of technologies
2. most trends follow a two phase pattern, new approaches are different from the existing, after 5 or two years an integration will happen.

In the abstract, he identifies six trends that he summarized as follows (numbers added by us):

(1) First, the physicality of interactions and the physical space of the learner became genuine components of digital education. The frontier between the digital and the physical has faded out. (2) Similarly, the opposition between individual and social views on cognition has been subsumed by integrated learning scenarios, which means that AIED pays more attention today to social interactions than it did at its outset. (3) Another trend is the processing of learners’ behavioural particles, which do not carry very many semantics when considered individually, but are predictive of knowledge states when large data sets are processed with machine learning methods. (4/5) The development of probabilistic models and the integration of crowdsourcing methods has produced another trend: the design of learning environments has become less deterministic than before. The notion of learning environment evolved from a rather closed box to an open ecosystem in which multiple components are distributed over multiple platforms and where multiple stakeholders interact.(6) Among these stakeholders, it is important to notice that teachers play a more important role than before: they interact not only at the design phase (authoring) but also in the runtime phase (orchestration). These trends are not specific to AIED; they depict the evolution of learning technologies as a whole.

Links

• The International Artificial Intelligence in Education Society. This organization sponsors a major conference every two years, e.g. AI&ED '05.

• The International Journal of Artificial Intelligence in Education (IJAIED), formerly Journal of Artificial Intelligence in Education (JAIE)

• Education page of the American Association of Artificial Intelligence]

• Journal of Interactive Learning Research (JILR)

• the Journal of Learning Sciences. An Official Publication of the International Society of the Learning Sciences

• International Journal of Computer-Supported Collaborative Learning. An Official Publication of the International Society of the Learning Sciences

References

• Anderson, J.R., Conrad, F.G. & Corbett, A.T. (1989) Skill aquisition and the Lisp Tutor. Cognitive Science, 13, 467-505.
• Burton, R.R. & Brown, J.S. (1982) An investigation of computer coaching for informal learning activities. In D. Sleeman & J.S. Brown (Eds), Intelligent Tutoring Systems (pp. 201-225). New York: Academic Press.
• Clancey, W.J. (1987) Knowledge-based tutoring: the Guidon Program. Cambridge, Massachusetts: MIT Press.
• Dillenbourg, P. (1994). The role of artificial intelligence techniques in training software, Paper presented at LEARNTEC 1994. PDF.
• Steels L (1990) Components of Expertise. AI Magazine, vol.11, 2,pp. 28-49.