SLIDES: https://docs.google.com/presentation/d/1wGNfJQv76-t29lOLs-MiyPxYhbfNGwtZUDQ0waWCb18/edit?usp=sharing Most dynamic theories of discourse include components that track the past of the discourse: the Common Ground (Stalnaker, 1978) tracks common beliefs established in the past, and Question Under Discussion (QUD) stack (Roberts, 1996/2012) tracks lines of inquiry that have been previously taken up. However, Roberts’s integrated model of discourse struggles with refusals to address the QUD as in (1). It incorrectly predicts such moves to be irrelevant (Hyska, 2015), and gives an ad hoc treatment of how these moves operate on the QUD stack. I argue that the dynamics of stack-based models of discourse can be repaired and simplified by directly modeling possible future discourse states in the context. I adopt a framework based on inquisitive semantics Ciardelli et al. (2018) and the commitment space model of Cohen and Krifka (2014) which lifts the discourse context to a set of information states, with each member representing a possible future state of the discourse. I give a pragmatic analysis of refusals to address a QUD (1) as well as a lexical semantics for the discourse particle *just* which rejects a QUD (1b). The proposal adds to a growing body of work arguing that the discourse context contains reference to not just the past, but also the future of the discourse (Cohen and Krifka, 2014; Onea, 2016; Warstadt, 2020). *(1) A: Why did Skp break up with you? * *a. B: I’d rather not say. * *b. B: They just did./#They did.*