While AI ethics interventions often focus on how researchers should navigate consequential choices, they may overlook a prior question: when do researchers recognize they are making a decision at all? This qualitative study examines how academic NLP teams confront “decision moments” – junctures where latent alternative paths could be considered. We propose a railyard problem analogy: where trolley problems presume a discrete choice between visible options, railyard problems concern whether alternative paths register as possibilities at all. Drawing on decision-tracing interviews across four NLP projects, we demonstrate how technical defaults, institutional structures, and tacit norms (infraethics) combine to organize research as a human-infrastructural process. Many consequential outcomes arise through "passive decisions", where alternatives exist but never become sufficiently visible, viable, or voiced (VVV) to warrant deliberation; "active decisions" only emerge when VVV conditions are met. Our analysis suggests ethics interventions should cultivate the collaborative conditions under which alternatives become recognizable.

LinkedIn is central to salaried job search and professional networking. In a career development program for adults seeking upward socioeconomic mobility through middle-wage computing work, we aimed to use LinkedIn to find and develop new social ties. However, we could not use the platform for this purpose. Through a participatory research approach, we formed a research team with diverse positionalities to understand why LinkedIn was difficult to use and how it could be better for our program. We analyzed recorded walk-throughs and confirmed our findings with two years of ethnographic field notes and written reflections. Our findings demonstrate that LinkedIn's embedded algorithms and interface design prioritize users with large networks who can afford a LinkedIn Premium subscription. We argue that such platform-embedded power differentials lead to platform-delivered microaggressions. Non-Premium users and users with small networks must endure microaggressions to participate in the salaried labor market. We argue the politics of LinkedIn as a platform are such that its embedded power differentials are beyond our control and unlikely to change. Therefore, we recommend sociotechnical coping and mitigation strategies for career development programs in lieu of design implications for LinkedIn or similar platforms. We contribute a detailed example of how a technology reinforces pre-existing privilege without users' knowledge.