Wellesley College

This paper presents Collage Collaborator, a human-AI co-creation system designed to support early-stage, embodied creative exploration in physical collage-making. The system combines an interactive web interface with a generative AI backend to enable rapid image generation, iterative variations across medium, color, and theme, automatic documentation of collage possibilities, and customization through user’s articulated concepts. It specifically addresses challenges faced by novice collagers, such as uncertainty around placement, imagery, and material choice, by offering a wide range of generative visual suggestions that users may adapt, reinterpret, or reject. Rather than producing finished artifacts, the system supports material decision-making while preserving creative agency and authorship. Findings from a preliminary evaluation suggest that this form of scaffolding promotes divergent exploration, material discovery, and creative confidence, while also surfacing tensions around feasibility and originality. Overall, this work demonstrates the potential of AI as an ideation partner that augments everyday creative practice by enabling experimentation beyond user’s technical skill levels.

Physiological signals are increasingly used to study stress, engagement, and user experience. Yet collecting and interpreting these signals remains difficult for students and novice researchers, who must navigate complex device setups, scattered software tools, and technical analysis workflows. This demo presents Sensei, an intelligent browser-based platform that streamlines the entire process of running small physiological studies—from connecting sensors and defining protocols to visualizing data and generating interpretations. In this demo, attendees will run a brief recording session and explore Sensei’s AI-assisted analysis tools, illustrating how intelligent support can make physiological computing more approachable for learning and early-stage research.

Sedimentary records retrieved from Walden Pond, Concord, MA and Sluice Pond, Lynn MA show evidence of mass wasting likely triggered by strong historic and prehistoric groundshaking. The two lakes of glacial origin form steep-sided basins of 30 m and 20 m maximum water depth with Walden Pond being divided into three distinct subbasins. Because of the visually homogeneous nature of the organic-rich lacustrine muds we employed a combination of sedimentological, geophysical, geochemical, and palynological proxies to characterize background sediments and the fingerprint of the M=5.9 1755 Cape Ann earthquake, the largest historic earthquake in New England. We then used a similar multiproxy approach to investigate the longer record of Sluice Pond as well as the individual records of the three basins of Walden Pond extending the paleoseismic record about 800 years back in time. Robust age models using a combination of radioisotopes, industrial contaminants, and pollen stratigraphy allow for dating of event horizons as well as correlation between different lake basins. Deposited before the arrival of Europeans settlers, sediments in both lakes are composed of organic-rich mud reflecting a forested landscape. The rise of non-arboreal pollen including Ambrosia record the widespread logging practices of European settlers in New England starting in the mid-17th century. Subsequent erosion of soils is marked by grain size decrease, lower organic matter content, and elemental compositions characteristic of local soils. Separate from these long-term trends we identified sediment horizons that mark sudden-onset events. These are typically characterized by larger grain size, influx of nearshore palynomorphs, higher bulk densities, and/ or lower organic matter content. We interpret these horizons as underwater mass wasting events. For Sluice Pond, we were able to identify two event horizons, one associated with the M=5.9 1755 Cape Ann earthquake, and another one dated to between 1390 -1510 CE. For Walden Pond, sediments show synchronous mass wasting in at least two subbasins potentially related to the historic 1755 earthquake, as well as two older events dated to between 1380-1520 CE and 1050-1320 CE. Interestingly, a cluster of landslides in southern Maine within geotechnically sensitive clays dates to 1220-1450 CE and could correlate with the disturbed lake sediments. We analyzed the spatial distribution of event horizons in the two lakes in Massachusetts and landslides in Maine and identified possible source areas and magnitudes of prehistoric earthquakes with one possible source area within the Littleton, MA seismic zone. Our analysis suggests that the seismic hazard in the Boston area might be higher than previously appreciated.

ABSTRACTThe connectivity and interdependence of our world and its inhabitants’ health have come under increasing focus. Elucidation of the common and interdependent mechanisms of health and disease requires approaches that facilitate understanding of complex systems behavior and probing of both individual and collective system parameters. To this end, multiscale physical and computational modeling offers a particularly powerful tool to predict behavior over vast time and length scales. Other novel technologies, e.g., rapid isolation nanotechnology developed to analyze nanoscale small extracellular vesicles in ocular tears, enable tracking of “fingerprints” from diseases as diverse as ocular to neurodegenerative (e.g., dementia) and cancer. In the future, it will be possible to track the health and disease of ecosystems and their inhabitants, using geospatial and epidemiological approaches, as well as novel biotechnologies, to prevent and mitigate disease processes and enhance well‐being. These concepts are applied by way of an exemplary approach to understand and address the impact of toxic, recalcitrant manmade chemicals (i.e., PFAS) on the health of ecosystems and their diverse inhabitants. Such convergent efforts will be necessary and a priority for solving the complex problems threatening the health of our planet and its inhabitants.