Abstract:

Digital media can particularly contribute to learners’ knowledge acquisition (Kerres, 2002) if digital media actively engage learners in interacting with the to-be-learnt subject area (keyword interactivity, Haak, 2002) as well as independently adapt to learners’ requirements such as prior knowledge and interests (keyword adaptivity, Petko & Reusser, 2006).
However, authors of digital media hardly integrate interactivity and adaptivity into their products due to the effort that is connected to the creation, for example, of an required items or task set. One option to reduce this effort is the use of a process that is called automatic item generation (AIG; Gierl, Lai, & Turner, 2012). In AIG, experts do not produce items by hand anymore. Rather, they create a systematic representation of the to-be-learnt subject area that a software (i.e. an item generator) uses to create a bunch of tasks or items.

In order to illustrate the AIG process, in the following, a short example is given: “In the subject area of network control, the TCP/IP-model can explain how data should be packetized, addressed, and transmitted for internet communication. The model itself consists of seven communication protocols (e.g., DHCP) as well as four layers (e.g., the application layer).” After the expert gathered this information in a systematic representation, the AIG-software can create 28 tasks or items by combining every protocol with every layer and then asking for a protocol with a given network layer or the other way around. To build more complex tasks we can add properties to every protocol. This will increase the number of generated questions and items.

We save the models in a web ontology language (OWL) representation. OWL is a common language for knowledge representations e.g. in biology or medicine. Thus we are able to import other models in our item generator. The generator component converts the OWL model to a graph model and computes all possible combinations of questions and items. At the moment we work on a web-based editor for the knowledge representations because common OWL editors are very complex and not easy to use.

However, as illustrated by this example, the complexity of the automatic generated items is relatively low up to now. That means that AIG-task can currently only address low-level learning goals such as the recall or recognition of to-be-learnt information. This is mainly due to the fact that AIG-software can only create tasks form unambiguous connections between information at present. Branching or loops in the systematic representation which may require learners to analyse the to-be-learnt material are currently not included. In a current project, we contributed to solving this problem.

References:
[1] Gierl, M. J.; Lai, H.; Turner, S. R.: Using automatic item generation to create multiple‐choice test items. Medical Education 46/8, S. 757-765, 2012.
Kerres, M.: Online- und Präsenzelemente in hybriden Lernarrangements kombinieren. In (Hohenstein, A.; Wilbers, K., Hrsg.): Handbuch E-Learning. Fachverlag Deutscher Wirtschaftsdienst, Köln, S. 1-19, 2002.
[2] Haak, J.: Interaktivität als Kennzeichen von Multimedia und Hypermedia. In (Issing, L. J.; Klimsa, P., Hrsg.): Information und Lernen mit Multimedia und Internet. Beltz, Weinheim, S. 127-136, 2002.
[3] Petko, D.; Reusser, K.: Das Potenzial interaktiver Lernressourcen zur Förderung von Lernprozessen. In (Miller, D., Hrsg.): E-Learning - Eine multiperspektivische Standortbestimmung. Haupt Verlag, Bern, S. 183-207, 2006.

Keywords:

Tasks, AIG, Items, Assessment, Interactivity, Adaptivity.