Add new post

Towards an argument interchange format, a proposal.

Towards an Argument Interchange Format

Carlos I. ChesñevarJarred McGinnisSanjay ModgilIyad RahwanChris ReedGuillermo R. SimariMatthew SouthGerard VreeswijkSteven WillmottThe Knowledge Engineering Review. Volume 21, Number 4, December 2006. pp. 293-316 (see


The theory of argumentation is a rich, interdisciplinary area of research straddling the fields of artificial intelligence, philosophy, communication studies, linguistics and psychology. In the last few years, significant progress has been made in understanding the theoretical properties of different argumentation logics. However, one major barrier to the development and practical deployment of argumentation systems is the lack of a shared, agreed notation or ‘interchange format’ for argumentation and arguments. In this paper, we describe a draft specification for an argument interchange format (AIF) intended for representation and exchange of data between various argumentation tools and agent-based applications. It represents a consensus ‘abstract model’ established by researchers across fields of argumentation, artificial intelligence and multi-agent systems1. In its current form, this specification is intended as a starting point for further discussion
and elaboration by the community, rather than an attempt at a definitive, all-encompassing model. However, to demonstrate proof of concept, a use case scenario is briefly described. Moreover, three concrete realizations or ‘reifications’ of the abstract model are illustrated.

Areas of accomodation

One issue that we face is accomodating ancillary functionality:

  • Whether to accomodate
  • How much
  • How

The last item, in my view, suggests a cautious answer:  In order to keep our scope small, perhaps we can proceed by merely surveying primary parties about what ancillary information they need to read and write, choosing a suitable location and existing format, and leaving it at that.


Potential areas wanting accommodation, not all of which we may wish to support, might include:

  • Mechanical theorem provers and proof checkers
    • Writing proofs into this format
    • Reading and proof-checking arguments that aspire to rigor.
    • Annotating this format with the results of such checking.
  • Argument visualization software
    • Reading the format and producing visualizations.
      • Expecting some sort of visual placement hints.
      • Or (better for us) deciding visual placement entirely by their own means.
  • Editors for transcribing or annotating arguments.  I honestly don’t know the state of the art here at all.
  • Automated transcription, such as by natural language understanding.

Please feel free to comment.

State of argument representation

First, I’m very much aware that there’s more going on in software and formats than I can personally keep track of.  So please, everybody, feel free to comment with corrections and updates.

That said, as I see it, the situation as we begin this group is this.  There are or have recently been several strong contenders for a standard format.  We are not bound to recommend one, modified or otherwise, but  for background I’ll list the popular ones to the best of my knowledge:

  • One formerly popular format, AML, has been cut loose by Araucaria.
  • AIF has become a popular representation format, probably the most popular at the moment.  It does a number of things right in my opinion.
  • LKIF is also fairly popular.  One issue with LKIF is that it is not primarily an argumentation format; argument representation is one module among many.  One the other hand, it has the virtue of having “work experience”, as it were, in representing contentious real world legal arguments.

“Statements or statement-like objects”

Our mission statement refers to “statement-like objects”. In order to
dispel any lingering mystery, I’ll talk about that in this post.

I phrase it that way so that we don’t commit prematurely to a
representation. A mission statement would be too early to decide that. Nevertheless, they are constrained to be statement-like and connected so that our concept of argument isn’t too anything-goes. I think that’s the right balance.

Statement-like objects could conceivably include, for instance:

  • Formulas that represent statements but arguably are not actually statements.
  • MECE sets of statements, should we choose to treat them as single objects.


Welcome to the Argument Representation Community Group.

Argument Representation’s mission is to recommend a standardized representation for formal argument.

The group does not neccessarily commit to creating a novel
representation. For instance, after due consideration it could
endorse an existing one or recommend accepting an existing one with
minor changes.

Formal argument is taken to mean a formalizable set of connected
statements or statement-like objects intended to establish a

Argument-Representation does not intend to augment XML in any other