This paper (44 page PDF) reviews more than two hundred samples "to explore the state-of-the-art in the annotation of behavior." It is a lot of work to read, and honestly, I couldn't devote the time to it that it deserves. That said, it is an excellent account of the sort of thinking needed in the development of educational technology in the future. What is it to perform a task? How do we describe it? What language do we even use? How do we even get people to expend the care and consideration needed to create a quality dataset that can be used productively by foundational models interpreting and supporting learning tasks? "There is a clear pattern towards the use of sophisticated semi-automatic and automatic methods annotation methods (however) he limiting factor in AI is shifting from a lack of capability to a lack of trustworthiness. (So) offering social annotation in the forms of peer dataset auditing and curation may... provide possibilities for the wider public to alleviate concerns through sublimating them into a productive and meaningful activity."
Today: 0 Total: 19 [Direct link] [Share]
