Conversational Agency in Augmentative and Alternative Communication
Stephanie Valencia, Amy Pavel, Jared Santa Maria, Seunga (Gloria) Yu, Jeffrey P. Bigham, Henny Admoni · 2020 · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376376
Summary
This paper introduces conversational agency as a new framework for studying how augmented communicators (ACs) — people who use AAC devices to speak — participate in and influence conversations. While prior AAC research has focused primarily on improving device throughput and efficiency, this study examines the broader social and interactional constraints that shape what ACs can say, when they can speak, and who they can address. The researchers studied four triads, each consisting of an expert AC with cerebral palsy (10+ years of AAC experience), their close conversation partner (CCP — a parent or paid aide), and an unfamiliar third party (the experimenter). Each triad completed three tasks: a semi-structured interview, a collaborative map-drawing task, and a craft collage task. The study draws on Gibson's theory of conversational agency, which identifies four core constraints: one-speaker (only one person speaks at a time), participation shift (who speaks next), relevance (contributions must relate to recent discussion), and ritual (maintaining social standing). The researchers developed quantitative metrics using video coding to measure both technical agency (participating in conversation) and colloquial agency (achieving conversational objectives).
Key findings
CCPs supported AC agency in three ways but also inadvertently constrained it. First, CCPs facilitated communication with third parties by interpreting device pronunciation, explaining AC gestures, and physically orienting conversations toward the AC. Second, CCPs increased information exchange through "delegated technical agency" — expanding on AC responses with permission or answering questions on the AC's behalf. Third, however, CCPs created "missed opportunities" by providing lengthy explanations that shifted conversation topics before ACs could contribute, or by guessing what ACs intended to say, limiting responses to yes/no confirmations. Device properties significantly impacted agency: ACs with utterance-level voice output (speaking complete words as typed) held the conversational floor better and participated more independently than those with phrase-level output (speaking only after completing entire phrases). Relationship type mattered too — parent CCPs participated more across all tasks than paid aide CCPs, and ACs used more nonverbal communication (gestures, vocalizations) with familiar partners. The relevance constraint disproportionately affected ACs: by the time they finished typing a response, the conversation had often moved on, forcing them to either erase their contribution or make an off-topic comment.
Relevance
This paper fundamentally reframes AAC research from a device-centric to a socially-situated perspective, which has important implications for how AAC technologies are designed and evaluated. For practitioners, the key insight is that improving device speed alone is insufficient — the social dynamics of conversation, the behaviors of communication partners, and the type of task all shape whether an AC can meaningfully participate. The concept of "missed opportunities" is particularly actionable: AAC systems could incorporate features like conversation bookmarking (saving a partially-typed response for later), status indicators visible to conversation partners (showing the AC is composing a message), or contextual word suggestions based on the ongoing conversation to help ACs keep pace. For organizations and caregivers, the findings highlight the tension between helpful facilitation and inadvertent silencing, suggesting that communication partner training should address not just how to support ACs but also when to step back.
Tags: AAC · augmentative and alternative communication · cerebral palsy · conversational agency · communication partners · assistive technology · speech generating devices · turn-taking · accessibility