Rose–Hulman Institute of Technology

To promote undergraduate research on Hollow Cathodes, Rose-Hulman Electric Propulsion Group developed a low-current Heaterless Hollow Cathode. This paper presents the design and testing of a low-current Heaterless Hollow Cathode. Differences to conventional cathode designs, and a detailed testing procedure will be covered. Furthermore, a characterization setup will be proposed. Ignition of the cathode was achieved. However, a lack of discharge current was observed during the experimental campaign. The cathode was observed to be operating in space charge-limited mode.

Optical trapping offers robust nanoscale control of matter but, to date, has been dominated by the interaction between a material’s electric polarizability, αe, and the electric part of light, therefore defined by electric-field intensity-gradient forces. Magnetic light-matter interactions, despite their potential to reshape optical trapping research, have remained experimentally unrealized. This paper addresses this long-standing deficiency by realizing optical magnetic field-associated trapping of high-index (i.e., Si) nanoparticles. Experiments, validated by our theoretical framework and Maxwell stress tensor calculations, reveal the essential role of a material’s magnetic polarizability, αm, and electric-magnetic scattering forces arising from the photonic Hall effect. This magnetic contribution allows exploration of stable trapping, distinct from purely electric-field control. Our findings open avenues for nanoparticle manipulation beyond conventional paradigms, enable previously unexamined optical matter formation driven by magnetic interactions, and suggest unexplored N-body effects and symmetry-breaking dynamics in optical matter systems.

This full innovative practice paper explores two concepts that can help build and maintain cohesion in faculty teams, namely psychological safety within a team, and capacity to navigate conflicts in a psychologically safe and productive way. Faculty teams working together through co-teaching models or through interdisciplinary research are becoming more prevalent, and inclusive and efficient teams are key to generating innovative, cross-cutting, and sustainable changes in higher education. Psychological safety of team members - defined as a shared belief held by team members that the other members of the team will not embarrass, reject, or punish them for speaking up - is key to enabling high performance in teams. However, even in the presence of psychological safety, conflict will occur in a team-setting. Teams that have built capacity to respond to that conflict productively can reestablish psychological safety and enable progress. We present a research informed faculty professional development module that can be utilized by faculty teams to practice more effective teaming, moving them away from the cognitive domain (awareness that teaming issues might exist) to the affective domain (being able and willing to act to improve the teaming situation). Specifically, the module includes reflective and role-playing activities to: (i) practice speaking and behaving in ways that foster psychological safety in teams, and (ii) practice having crucial conversations in instances when psychological safety is eroded. This innovative module is subject matter agnostic and highly transferable and therefore will be of value to any faculty member working in a team setting.

This paper addresses the issues of parameter redundancy, rigid structure, and limited task adaptability in the fine-tuning of large language models. It proposes an adapter-based fine-tuning method built on a structure-learnable mechanism. By introducing differentiable gating functions and structural sparsity control variables, the method enables automatic optimization of adapter insertion points, activation paths, and module combinations. This allows the model to adjust its structure flexibly in multi-task settings to match different task characteristics. With the backbone parameters kept frozen, the method uses a structure search mechanism to guide the dynamic construction of task-specific efficient substructures during training. This significantly improves parameter utilization and representational capacity. In addition, the paper designs a set of sensitivity analysis experiments to systematically evaluate the effects of sparsity weight, noise injection ratio, and data perturbation on model performance. These experiments verify the stability and robustness of the proposed method across various multi-task natural language understanding tasks. The experimental results show that the proposed method outperforms mainstream parameter-efficient tuning techniques on multiple tasks. It achieves a better balance among accuracy, compression rate, and robustness to noise and perturbation.

Abstract Gerrymandering—the deliberate manipulation of electoral district boundaries for political advantage—is a persistent challenge in U.S. elections. In this work, we introduce and analyze Votemandering, a strategic blend of gerrymandering and targeted political campaigning devised to gain more seats by circumventing fairness measures. Votemandering leverages accurate demographic and socio-political data, bolstered by advancements in technology and data analytics, to influence voter decisions in pursuit of subtle gerrymandering strategies. We formulate votemandering as a Mixed Integer Program (MIP) that performs fairness-constrained gerrymandering over multiple election rounds. We analyze the influence of various redistricting constraints and parameters on votemandering efficacy. We explore the interconnectedness of gerrymandering, substantial campaign budgets, and strategic campaigning, illustrating their collective potential to generate biased electoral maps. A case study of Wisconsin State Senate redistricting reveals significant votemandering potential. Our findings underscore the need for reforms in the redistricting process beyond enforcing thresholds for specific fairness metrics.

Abstract In this paper we present two modifications of traditional NK fitness landscapes, the θNK and HNK models, and explore these modifications via accessibility and ruggedness. The θNK model introduces a parameter θ to integrate local Rough Mount Fuji-type correlations in subgenotype contributions, simulating more biologically realistic correlated fitness effects. The HNK model incorporates gene regulation effects by introducing a masking mechanism where certain loci modulate the expression of other loci, simulating effects observed in gene regulatory networks without modeling the full network. Through extensive simulations across a wide range of parameters (N, K, θ, and H), we analyze the impact of these modifications on landscape accessibility and the number of local optima. We find that increasing θ or the number of masking loci (H) generally enhances accessibility, even in landscapes with many local optima, showing that ruggedness doesn’t necessarily hinder evolutionary pathways. Additionally, distinct interaction patterns (blocked, adjacent, random) lead to different observations in accessibility and optimum structure. While more complex than traditional NK, we believe each model provides a new biologically relevant facet to fitness landscapes and provides insight into how genetic and regulatory structures influence the evolutionary potential of populations.

This poster presents a novel methodology to guide software engineers in developing ethically grounded software systems, ensuring that they proactively avoid ethical pitfalls such as bias and privacy violations. The proposed methodology is based on the concept of “deconstruction” originally conceptualized by philosopher Jacques Derrida [1]–[4]. It emphasizes a comprehensive approach to addressing ethical dilemmas, particularly in managing opposition within ethical disputes, where conflicting perspectives often arise.

Abstract Public statements by leading AI researchers and recognizable people in the computer world are suggesting that AI may soon replace many jobs, including software engineers. Some even state that soon, AI will be smarter than us. We believe such statements are unhelpful when it comes to attracting talent to our field. We document several such statements. We believe that the future need for AI talent is tremendous and that we should take extreme efforts to attract students to our field. We present a sample of the expected opportunities and needs. Some of these opportunities may be attractive to students who in the past may not have considered AI as a career option. We argue that even with the anticipated automation of AI work, there nevertheless will be a prodigious need for talent to develop good AI. We summarize work that argues that AI is going to be a fundamental skill and as such should be introduced to learners across many age groups and many backgrounds. We suggest that a well‐reasoned statement of the anticipated needs be developed by experts in our field and communicated to future talent. We suggest that, as part of this message, pathways forward toward developing AI talent across a wide range of backgrounds be developed and communicated.