Search

Refereed/Peer-reviewed The research objective of this paper is to advance knowledge about the role of artificial intelligence (AI) in complex problem-solving. A problem is complex due to the large number of highly inter-connected variables affecting the problem state. Complex problem-solving situations often change decremental or worsen, forcing a problem solver to act immediately, under considerable time pressure. While research findings support the assumption that (1) affective, (2) (meta-)cognitive, and (3) social processes support complex problem-solving, opportunities of AI for supporting complex problem-solving need to be further investigated. This article presents a scoping review of relevant literature from the last five years. The study included, N = 38 studies for coding and analysis. Our findings show that in addition to the increased number of publications, the current trend suggests increased quality of published work. Human-AI collaboration in complex problem-solving has been explored across a broad variety of AI application domains. However, the four dimensions of complex problem – namely, cognitive, metacognitive, social and affective - have been augmented by AI to a different extent. Although most of the work has been done in the cognitive domain, it is encouraging to see progress across social and affective dimensions as well. Implications for future research and practice are being discussed.

usc Predictions of the timelines for when machines will be able to perform general cognitive activities that rival humans, or even the arrival of “super intelligence”, range from years to decades to never. For researchers in the education sector, the potential future state of AI, while provocative, is secondary to important shorter-term questions that influence how AI is integrated into learning and knowledge practices such as sensemaking and decision making. AI is not a future technology. It is already present in our daily lives, often shaping, behind the scenes, the types of information we encounter. It is, therefore, important to consider immediate questions surrounding the dynamics of human-machine interactions. In this paper, we focus on the relationship between human and artificial cognition and treat these as separate systems, each with distinct strengths and capabilities. We adopt a functional view (i.e., discrete tasks) of the activities that artificial cognition completes and those that are best handled by humans. This creates a foundation to then evaluate models for how these two cognitive systems interact and the mechanisms for coordination that are required. In doing so, we create a basis for future researchers to develop testable hypotheses regarding the impact of artificial cognition on knowledge processes such as learning, sensemaking, and decision making. Our evaluation provides insight for researchers regarding the optimal relationship between which cognitive activities should be handed off to the machine, which should remain the domain of human performance, and how these two should then be integrated when outputs are passed from one cognitive system (human or artificial) to the other. Refereed/Peer-reviewed

Massive Open Online Courses have provided researchers with considerable opportunities to assess student learning within technically mediated online environments. However, for most reported studies, analyses are conducted from a single MOOC or over multiple MOOCs with different learner enrolments. Thus, this limits the opportunities to assess the changing behavior of learners over time. In this paper, the behavioral engagement of 175 students was examined, who were enrolled in a professional development study program consisting of four different MOOCs, comprising a complete program of study. A comprehensive analysis was conducted to understand the changing behavior of learners within each course and across the MOOC program. To do this, we used latent class analysis, a clustering technique widely used in other fields, but mostly unexplored within educational technology research. Our results revealed six weekly learning strategies based on student course engagement and three different program-level learning strategies, which not only differed in their learning behavior but also their final academic outcomes. Our study also showed substantial effects of MOOC course design on the level of student engagement in the courses. The results and implications are further discussed. Refereed/Peer-reviewed