How do we assess learning in complex technology-mediated practices? How does the coordination of technological af-fordances and constraints mediate immediate performance and individual learning? In the technology-mediated practiceof programming, the compiler functions as a source of both affordances and constraints to the human cognitive agent.The compiler affords the compilation of executable programs and dynamically informative compiler feedback, while thecompiler also constrains acceptable code to a specific syntax. In this in-class experimental study, I investigate the contri-bution of compiler affordances and constraints to performance and learning in programming. The study results indicateaffordances as important facilitators of immediate performance. Conversely, constraints appear important mediators ofconceptual learning, which in turn facilitates internalized thinking decoupled from the original technological resource.The findings imply a need for teaching and learning activities that emphasize practicing resource-coordination and foran assessment practice that intelligently combines technology-mediated resource-rich tasks with decoupled resource-poortasks.

Models of cognition and learning structure and inform the thinking and action of educational practitioners and researchersalike. They serve as a communication device both within and between research and practice. There is a need for a holis-tic framework of cognition that appropriately reflects and synthesizes the current state of the field of the cognitive andlearning sciences with its rich diversity of research agendas. I propose such a model, which conceptualizes learning asunfolding from three interlinked basic domains: Conscious thought in the form of percepts and symbolic representationsin a symbolic-conceptual domain; foundational preconscious processing in a domain of cognitive metaphor; as well assituated, embodied interaction in a tangible enculturated agent-environment domain. The fundamental theoretical com-mitment of this Holistic Framework of Cognition and Learning is to dynamical systems theory. Emergence serves as thefunctional binder that ties the frameworks seemingly disparate elements together into a coherent whole.

Interaction with environmental resources such as technological tools facilitates conceptual learning and preparation for future learning. Writing and compiling code, failing, and trying again fosters deep understanding of computational concepts. This study investigates knowledge gap awareness as potential mediating cognitive mechanism in the learning process from situated interaction with a technological tool to internalized conceptual understanding. Students engaged in a tool-mediated or tool-dissociated learning activity on functional programming and were subsequently assessed on their learning again under tool-mediated or tool-dissociated conditions. Students’ assessment performances were triangulated with their open comments and questionnaire responses on the study’s learning and assessment activities. The findings support the assumption of tool-mediation facilitating knowledge gap awareness, which in turn facilitates conceptual learning.

Outside university, students encounter disciplinary practices mediated by technological resources. In this sense, the real world is decidedly resource-rich. In contrast, most educational assessments remain decidedly resource-poor. Situated versus mindbased perspectives of cognition fundamentally differ in the role they ascribe to such resources in cognition and learning. To mindbased perspectives, they are a source of input, to situated perspectives they are constitutive to cognition itself. We assessed the validity of resource-rich versus resource-poor assessments of learning outcomes from resource-rich versus resource-poor learning activities. The study implemented an in-class 2x2 between-subjects experimental design in an introductory programming course with 192 first semester BSc engineering students. Both types of assessment were sensitive to differences in learning outcomes, indicating validity for both. Results indicate resource-rich assessments may be more ecologically valid, while – intriguingly – the resource-poor assessments were more sensitive to transfer of learning. Furthermore, the resource-rich learning activities better facilitated learning for transfer.