May 14, 2014 | Written by: Satya Nitta
The recent emergence of highly interactive technologies such as touch technologies, natural language based interactions with computing, gesture recognition, visual immersion and the maturing of speech recognition are historically important milestones in computing. Together, these technologies, along with other advances in the field of artificial intelligence (AI), are ushering in the era of cognitive computing, where the computers have attributes that allow them to learn and interact with humans in more natural ways. At the same time, advances in neuroscience, driven in part by progress in using supercomputers to model aspects of the brain as well, as by projects such as the human connectome project promise to bring us closer to a deeper understanding of some cognitive processes such as learning. At the intersection of cognitive neuroscience and cognitive computing lies an extraordinary opportunity. This opportunity could allow us to refine cognitive theories of learning as well as derive new principles that should guide how learning content should be structured when using cognitive computing based technologies. The result could, over time, completely transform digital learning content and make learning a much more engaging process, tuned to generation that is coming of age in these extraordinary times.
At IBM Research, we call such interactive content “Cognitive Learning Content” and have embarked on the first steps towards creating working prototypes to help develop this field through partnerships with like minded institutions. To do so, we are leveraging Watson, the US game show winning computer. Watson, is an example of a cognitive computer. Watson displays a remarkable understanding of natural language and allows the user to interact with it through natural language based queries. An introduction to the technical architecture and process pipeline for the Watson system can be found at this reference (1). Further, aspects under current development with Watson such as dialog and visualizing inference pathways allow for a much deeper level of interaction between the user and the machine. In addition to Watson based interactivity, using other modes of interaction such as gesture, touch, as well as speech recognition, integrated with personalized adaptive learning systems are being considered. Aspects such as gamification and highly visual ways of representing and interacting with learning content are also central to this initiative.
There are two critical considerations to designing content with this level of interactivity. The first, as noted above, is that content such as this has to be designed with a deep understanding of underlying cognitive neuroscience as well as cognitive theories of learning. The second is that there needs to be a closed feedback loop that assesses, informs and refreshes the principles used to design the content to ensure it’s efficacy.
Initial use-cases of such highly interactive, intelligent content revolve around notions of a cognitive tutor to assist learners as well as educators. These tutors become especially powerful when integrated with personalized adaptive learning systems. Aspects such as automated scoring of constructed responses as well as automatically generated suggestions for remedial content will all be expected to feature in such tutors. Other use-cases revolve around assisting content search, assessing content accuracy etc.
Ultimately, the vision for this initiative is to usher in not only the era of personalized learning, but also to an era where deeply immersive, interactive and intelligent content which is carefully designed to engage learners and drive better outcomes is deployed to the benefit of learners everywhere.
1. D. Ferrucci, IBM J of R&D V 56, No. ¾, May/July 2012