AI-powered educational technologies have demonstrated measurable benefits for learners, but their design and evaluation have largely centered on K-12 contexts. As a result, many AI-supported learning systems remain poorly aligned with the needs, constraints, and goals of adult learners. To better understand how AI systems function in adult education, this paper examines the deployment of several AI learning technologies developed within a multidisciplinary, national research institute in the United States focused on adult learning and online education. Drawing on longitudinal deployment data, we conducted a reflexive thematic analysis to identify recurring challenges and design considerations across systems. These insights were synthesized into a set of 19 design guidelines intended to inform future AI-supported adult learning technologies. We demonstrate the utility of these guidelines through a heuristic evaluation of the deployed systems. Lastly, we present a guideline exploration tool that aids in the ideation of technologies by connecting the guidelines to stakeholder statements surfaced in the analysis process.

A decade ago, massively open online courses (MOOCs) were heralded as the solution to universal global access to higher education. While they failed to reach this vision, primarily because of presentational/assimilative instruction (a PDF of the residential classroom), MOOCs provided foundational models and infrastructure for emergency remote learning in the pandemic. Reports of remote learning's death post-pandemic are greatly exaggerated, since the world is now irreversibly hybrid-and will stay that way because many people and organizations value the new opportunities this presents. From now on, all of us must have skills and attributes for adept, engaging performance both face-to-face and across distance. Moreover, online/hybrid interactions enable the multi-dimensional streams of data and natural language essential for personalizing instruction and motivating learning.

Research suggests that the socioeconomic status (SES) of students plays a role in teacher perceptions, but it is unclear if virtual simulation technologies can detect this bias. In this exploratory study, a convenience sample of former and current teachers was asked to evaluate students’ performance in a virtually simulated classroom environment. Teachers using the virtual classroom exhibited biased judgements when shown students from varying SES backgrounds who made equivalent progress on a literacy task. Most teachers judged low SES students as making more progress than high SES students. Teachers reported feeling immersed in the experience, and immersion was correlated with bias in teacher judgements. This underscores the value of immersive simulations as a mechanism for inducing realistic situations that draw out unconscious biases. This approach may be of particular value when teachers must recognize that they have biases, as a necessary step in unlearning these by practicing unbiased strategies in simulated environments, with the aim of transfer to real world classrooms.

Computational thinking and scientific modeling are essential STEM education practices, and visual block-based programming interfaces offer new opportunities to make coding accessible in elementary school science. The EcoMOD research project blends computational modeling and ecosystem science learning through a 3D virtual forest ecosystem combined with a 2D visual block-based programming tool, for 3rd and 4th grade students (ages 8-10). During a 14-day curriculum, students explore an immersive virtual forest ecosystem, and program agent-based computational models of a beaver building a dam. As students run their program and watch their beaver agent go through the steps of building a dam, they observe the emergent outcomes as their programmed model impacts other elements of the ecosystem. The final curriculum was implemented with 7 teachers and approximately 150 students.

To scaffold students' investigations of an inquiry-based immersive virtual world for science education without undercutting the affordances an open-ended activity provides, this study explores ways time-stamped log files of groups' actions may enable the automatic generation of formative supports. Groups' logged actions in the virtual world are filtered via principal component analysis to provide a time series trajectory showing the rate of their investigative activities over time. This technique functions well in open-ended environments and examines the entire course of their experience in the virtual world instead of specific subsequences. Groups' trajectories are grouped via k-means clustering to identify different typical pathways taken through the immersive virtual world. These different approaches are then correlated with learning gains across several survey constructs (affective dimensions, ecosystem science content, understanding of causality, and experimental methods) to see how various trends are associated with different outcomes. Differences by teacher and school are explored to see how best to support inclusion and success of a diverse array of learners.

In recent years, the field of education has challenged researchers and practitioners to incorporate computing as an essential focus of K12 STEM education. Widely recognized as a "basic skill" necessary for economic opportunity and social mobility, integrating computing within K12 STEM supports learners of all ages in applying computational thinking within established and accessible contexts while co-developing practices essential to mathematical and scientific expertise. The EcoMOD project is an example of such an integration. In EcoMOD, immersive virtual environments are blended with interactive computational modeling tools designed specifically for younger children in an attempt to support ecological knowledge, complex causal reasoning and computational creation in third grade students. Our curricular design is grounded in the Science-as-Practice perspective and literature on young children's computing. In this proposal, we present outcomes from early pilot work with 35 third graders across the greater Boston area. This work has implications for designers of educational technologies by illuminating aspects of children's understanding that researchers and designers can leverage to support deeper and earlier understandings of ecological dynamics and computational practices.

The invention of the movie camera was initially seen as a way to reach "massive" by filming plays; we are in an equivalent stage with our early MOOCs. Thinking outside the box of "teaching" is essential to realizing learning at scale. Virtual worlds and augmented realities can complement digitized classroom instruction through simulated apprenticeships, embedded support for learning everywhere, and transformed social interactions. Going big also requires thinking small: analyzing diagnostic micro-patterns to customize individual learning, sifting through millions of participants to find the ideal partners to aid each other's growth. To reach massive with universal access and powerful outcomes, we must creatively expand our visions of platforms, pedagogy, and financing.

Summary form only given, as follows. Over the next decade, three complementary interfaces will shape how people learn: (1) the familiar webpages-and-windows "desktop," (2) multi-user virtual environments (including sensory immersion via virtual reality), and (3) augmented realities based on mobile wireless devices and infused in real world settings. The "millennial" learning styles ascribed to the Net Generation stem primarily from the desktop interface; however, the growing use of virtual environments and augmented realities is fostering new forms of mediated learning in users of all ages. The crucial factor leading to this "neomillennial" learning is that the desktop interface is not psychologically immersive, while in contrast virtual environments and augmented realities induce a strong sense of "presence." Psychological immersion enables a powerful pedagogy, situated learning, which is based on authentic contexts, activities, and assessment coupled with guidance from expert mentoring as well as tacit learning through collaborative activities. Through situated learning enabled by psychological presence, virtual environments and augmented realities is shaping participants' learning styles beyond what using sophisticated computers and telecom-munications has fostered thus far, with multiple implications for education and training.

This paper describes a research study that investigated how designers can use frames of reference (egocentric, exocentric, and a combination of the two) to support the mastery of abstract multidimensional information. The primary focus of this study was the relationship between FORs and mastery; the secondary focus was on other factors (individual characteristics and interaction experience) that were likely to influence the relationship between FORs and mastery. This studys outcomes (1) clarify how FORs work in conjunction with other factors in shaping mastery, (2) highlight strengths and weaknesses of different FORs, (3) demonstrate the benefits of providing multiple FORs, and (4) provide the basis for our recommendations to HCI researchers and designers.