Project Description
Learning by scaffolding is an advantageous training approach since the learners can actively participate and clearly see their progress. Another advantage is that the learner is able to receive individualized treatment based on his/her training needs. This individualization, however, is also perhaps the biggest problem for the teacher since developing personalized supports and scaffolded lessons would be significantly time-consuming. Furthermore, this problem is exacerbated if you consider a large number of trainees across multiple training scenarios. Another problem occurs considering the personality of the teacher. A proper scaffolding session requires that the teacher give up some of the control and allow the students to make errors. Some conscientious teachers may find this difficult to do effectively. As a final problem, traditional manuals and guides in a learning environment do not include scaffolding instruction although the notion of scaffolding is independent of the actual training material.
We intend to investigate and exploit intelligent software mechanisms or intelligent agents to help mitigate the impact of the aforementioned problems. Agents are software entities that have the knowledge of their environment and the innate capability to learn and adapt to their context given external stimuli. Agents are particularly well equipped in this domain because their operations are based on rules. As a learner performs within a scaffolded routine, agents can observe response times and errors and automatically reconfigure training routines to the respond to the individual learner. In addition, agents do not have the burden of human emotions. An agent can consistently mandate a training routine without the barriers imposed by empathy (as with human instructors) when learners make mistakes.
Typical training evaluation materials have problem sets combined with the corresponding hints. Without reconstructing this evaluative material, intelligent agents can control the delivery of hints and ultimately transform traditional training material into a scaffolded approach. To deliver these hints, we propose a scaffolding agent that controls hints over multimodal interfaces (i.e. graphical, textual and audible). In addition, we believe that by having multiple agents controlling various training tasks that an organization-wide scaffolding profile can be created.
|
