Conversational scoreboard
In linguistics and philosophy of language, the conversational scoreboard is a theoretical representation of the state of a conversation at a given moment. It treats discourse as a kind of game in which each speech act updates a structured collection of contextual parameters – for example the common ground, the questions currently under discussion and the interlocutors' public commitments – thereby constraining which subsequent moves are appropriate or felicitous.
The notion was introduced by David Lewis in his paper "Scorekeeping in a Language Game", which compares conversation to a game of baseball whose scoreboard records the evolving status of play. Since then, variants of the scoreboard framework have played an important role in dynamic semantics, formal pragmatics and recent work in social and political philosophy of language.
History
Common ground and early context models
Lewis' scorekeeping model builds on earlier work by Robert Stalnaker on conversational context and common ground. Stalnaker characterises conversation as proceeding against an ever-changing background of propositions that interlocutors mutually accept, the common ground, whose intersection is a set of possible worlds compatible with what is taken for granted in the conversation. This common-ground view provided a template for treating context as an explicit object in semantic and pragmatic theorising.During the 1970s and 1980s, dynamic approaches to meaning in natural language semantics – including context-change semantics and file-change semantics – developed the idea that the meaning of an expression should be understood in terms of its potential to update such contextual representations. These developments set the stage for treating context as an information state that can be formally modelled and manipulated.
Lewis's scorekeeping model
In "Scorekeeping in a Language Game", Lewis proposes that a conversation can be modelled by a conversational score, a tuple of abstract parameters that determine the truth, appropriateness or acceptability of utterances in that conversation. The score includes, among other things, sets of presupposed propositions, conversationally salient restrictions on quantifiers, standards of precision for vague predicates and other parameters relevant to evaluating speech acts.Lewis emphasises two ideas. First, which utterances count as correct or acceptable depends on the current score, just as the legality of a move in baseball depends on the current count of balls, strikes and outs. Second, the score itself is affected by what speakers say, often via accommodation: the score may adjust so that an apparently acceptable utterance comes out true or felicitous, provided no interlocutor objects. This helps explain how presuppositions, contextual standards and other parameters can shift in the course of conversation.
Lewis' paper became a central reference point for later work on presupposition, indexicality, vagueness, modality and other context-sensitive phenomena in both philosophy and linguistics.
Subsequent developments
Later authors extended and modified Lewis' framework in several directions. Stalnaker refined his account of common ground and its interaction with assertion, arguing that making an assertion is characteristically a proposal to add its content to the common ground, subject to uptake by other interlocutors.Dynamic semantics and dynamic pragmatics generalised the idea of a conversational score as an evolving information state. Work by Jonathan Ginzburg on dialogue gameboards, for instance, models the context of a conversation using a structured record containing shared information, outstanding questions and other discourse-relevant parameters.
In formal pragmatics, Donka Farkas and Kim Bruce developed a context structure for modelling speech acts as updates to a score containing a common ground, individual discourse commitments, and a "Table" of issues currently under negotiation. In epistemology, Keith DeRose drew on Lewis to formulate "single scoreboard semantics" for epistemic contextualism, treating the truth conditions of knowledge ascriptions as determined by a single conversational score for a given exchange.
Recent surveys, such as the overview by Lars Dänzer, Stefan Rinner and Eugenia Kulakova, present "conversational scorekeeping" as a broad research programme encompassing these and other developments in semantics, pragmatics and philosophy of language.
Structure of the conversational scoreboard
There is no universally accepted list of components that every conversational scoreboard must contain, and different authors choose different parameters depending on their theoretical goals. However, many frameworks explicitly or implicitly treat the score as a tuple of abstract objects representing at least the following types of information:- Common ground and context set. A set of propositions that participants treat as mutually accepted, often represented by the set of possible worlds in which all of those propositions are true.
- Questions under discussion or issues on the Table. Many models posit an ordered set or stack of matters currently being negotiated, sometimes called the "Table" or the set of questions under discussion. This component structures conversational relevance and helps explain the role of focus and information structure.
- Discourse commitments. Some theories track, for each participant, the propositions that they have publicly committed themselves to during the conversation, whether or not these are yet part of the common ground.
- Discourse referents and information state. In many dynamic semantic frameworks, the score includes a stock of discourse referents – abstract markers for individuals, times or events introduced in the conversation – which serve as potential antecedents for anaphoric expressions such as pronouns.
- Normative and contextual parameters. Following Lewis, many authors assume additional parameters representing, for example, standards of precision for vague predicates, contextually relevant alternatives, ordering sources for modals or contextually salient comparison classes.
Score update and accommodation
On scorekeeping accounts, the meaning of a sentence is typically characterised in terms of how it updates the scoreboard. Assertions, questions, imperatives and other speech acts have characteristic update rules specifying how the various components of the score change when such an act is successfully performed.A central phenomenon for scorekeeping theories is accommodation. An utterance may presuppose that the score already contains certain information. If the utterance appears acceptable and no interlocutor challenges the presupposition, the score is adjusted so that the required information is added to the common ground, making the utterance evaluable as true or false. Lewis and later authors have argued that similar accommodation mechanisms operate for other score components, such as standards of precision or contextually relevant alternatives, helping to explain how conversational moves can reshape the space of permissible future moves.
Scorekeeping formalisms also allow theorists to model more fine-grained conversational dynamics, such as the difference between merely proposing an update and its actual acceptance by other participants, or the ways in which rejections, corrections and denials can revert or block proposed changes to the score.
Applications
Formal semantics and pragmatics
Scorekeeping has been applied to a variety of phenomena in formal semantics and pragmatics. Farkas and Bruce use a score-based context structure to analyse similarities and differences between assertions and polar questions, and to model how answers and response particles update the common ground, the Table and individual discourse commitments.Sophia Malamud and Tamina Stephenson employ a conversational scoreboard to analyse declarative force modifiers – expressions that affect a speaker's degree or type of commitment to the content of an utterance, such as epistemic adverbs and parenthetical clauses. In their account, such modifiers alter how an utterance interacts with the score, for example by preventing its content from being fully added to the common ground while still updating other components of the scoreboard.
Sunwoo Jeong uses a scoreboard-based framework to investigate intonation and sentence-type conventions, arguing that different kinds of rising declaratives correspond to distinct patterns of score update and commitments, distinct from both canonical assertions and questions. Other authors have extended scorekeeping approaches to discourse particles, evidentials and right-peripheral tags in a range of languages.
Computational and dialogical models
In computational linguistics and dialogue theory, scoreboard-like structures underlie many models of conversational agents and dialogue management. Jonathan Ginzburg's dialogue gameboard framework represents the state of a conversation using records that encode shared assumptions, current issues, pending moves and discourse referents, and specifies how these records are updated by various types of speech acts. Raquel Fernández and others have provided formal implementations of such "gameboards" using dynamic logic and related tools.These approaches connect the philosophical idea of scorekeeping with practical concerns in computational linguistics, such as reference resolution, clarification requests and the management of multi-party dialogue.