CFP: 2nd Call for papers “WRAICogS1 – Writing Aids at the Crossroads of AI, Cognitive Science, and NLP” (COLING workshop)

Posted on by

Second call for papers:

"WRAICogS1 – Writing Aids at the Crossroads of AI, Cognitive Science, and NLP"

KEYNOTE SPEAKER

Cerstin Mahlow, Professor of Digital Linguistics and Writing Research,

ZHAW School of Applied Linguistics, Winterthur, Switzerland

MOTIVATION

This workshop is dedicated to developing writing aids grounded in human cognition (limitations of attention and memory, typically observed habits, knowledge states, and information needs). In other words, we focus on the cognitive and engineering aspects of interactive writing. Our goal is not only to help people acquire and improve their writing skills but also to enhance their productivity. By leveraging computer technology, we aim to enable them to produce better texts in less time.

Writing is one of the four cornerstones of communication. By leaving a trace, it allows us to reach many people, to transcend space and time, and to spare ourselves the trouble of memorization. Writing is undeniably important, whether as a communication tool, a thinking aid, or a memorial support. However, what is less obvious is the process—that is, the precise steps required to transform an intuition or vague idea into concrete, well-polished prose. Producing readable, well-written text requires many skills, deep and broad knowledge of various sorts (topic, language, audience, metaknowledge, i.e., how to use the information at hand?)— a lot of practice and appropriate feedback.

No one can learn all this overnight. The quantity and diversity of knowledge to interiorize, as well as the variety of cognitive states encountered, may explain why writing is so difficult and why it takes time to gain control over the whole process and become an expert writer. Unfortunately, knowledge alone is not enough. Writing is also a time- and energy-consuming endeavor. It is very hard work.

Since writing is difficult, and since there are now computer programs capable of doing it, one may wonder:

– whether we should leave the job entirely to the machine, or

– whether we could use these programs to help people write or to acquire the skill of writing.

Indeed, there are situations where it makes sense to rely on machines (e.g., routine work, business letters), but there are also many situations where this strategy is not recommended (e.g., writing to understand, writing to enrich and clarify our thoughts, writing to support thinking). That being said, one may find a middle ground where humans and machines work together, each contributing their strengths. It remains to be seen where machines can assist in the process (e.g., idea generation, idea structuring, translation into language, revision, editing) and where it is better to leave control to humans. Hence, the main question is not whether we should use LLMs to produce texts, but rather how, when, and at what level to use them or other techniques to help people produce written text.

In sum, our main goal is not to substitute machines for people or to have them do the job in people’s place, but rather to have machines assist people. Specifically, we aim to help people learn to write, speed up the process, gain better control, and reduce stress and cognitive load. Our motivation is largely practical and educational.

Obviously, we are not the first ones to pursue this goal. However, while many workshops focused on developing educational software, creating intelligent writing assistants, or evaluating written text, the submitted papers have primarily addressed formal aspects, such as grammatical error detection and spotting spelling mistakes. Yet good writing (text composition) requires much more than just the production of well-formed sentences.

Our mission is to go beyond merely identifying errors or mistakes made at the very end of the writing process, such as those due to ignorance or inattention. Instead, we aim to evaluate the quality of the choices made at higher levels. In other words, we are interested in the full spectrum of writing, including technology-based writing aids that address all tasks involved in writing: conceptual planning (ideation, organization), linguistic expression, editing, and revision. Hence, we welcome papers that focus on the higher levels of composition—such as thinking, reasoning, and planning (idea generation, outline planning)—as well as those concerned with the lower levels (grammar, spelling, and punctuation).

Arguably, this is the first workshop to:

– Consider the entire spectrum of writing rather than only the lower levels,

– Integrate humans right from the start into the development cycle of writing aids, and

– Provide support and feedback at any moment —before, during, and after writing— rather than only at the very end.

TOPICS

We welcome contributions on all topics related to writing aids, including but not limited to the following:

1. THE HUMAN PERSPECTIVE: Cognitive scientific viewpoints, including education, psycholinguistics, and neuroscience.

(a) Support: How can AI tools support critical thinking and logical reasoning in writing? How can writing assistants tailor feedback to individual writers, considering their unique needs and styles? How can we assess the quality and impact of AI-generated feedback on students’ writing (methods, metrics, etc.)?

(b) Topical coherence: How can we help people organize their ideas into a coherent whole? How do we model or operationalize the concept of a topic, the paragraph’s most central element? How do we detect possible topics within our data? What are typical subtopics of a given topic, and how do we identify them? How do we cluster content/ideas into topics and give the clusters appropriate names?

(c) Building software: How do we include humans in the development cycle of writing aids? How and at what level can engineers use insights from psycholinguistics and neuroscience? How can they model the writing process while accounting for human and technological factors?

(d) Metacognition: What do people typically know about writing in general and their own writing in particular? What are their problems and needs? How do people manage to coordinate the different processes? What should an authoring ecosystem look like (components)? What could be automated, and what is best left for interactive processing?

(e) Shared tasks: What kinds of shared task would be meaningful while being technically feasible?

2. THE ENGINEERING SIDE

(a) LLMs: Where in the writing process could we use methods developed in AI (e.g., LLMs) or computational linguistics (e.g., content generation, content structuring, translation into language, revision)? What are the potential benefits, dangers, and limitations of LLMs as writing aids? How could revealing the ‘knowledge’ embedded within black-box models improve their effectiveness, particularly in terms of increasing the accuracy and relevance of the feedback they provide? How can we address challenges related to data collection, privacy, and ethical considerations in developing and deploying AI writing tools?

(b) Tools and resources: What kinds of tools and resources (e.g., Sketch Engine, Rhetorical Structure Theory, knowledge graphs, and linked data) could be useful?

(c) Quality assessment: How can we check the veracity of facts, relevance, cohesion, coherence, style, fluency, proper use of pronouns, grammar, word choice, spelling, and punctuation?

(d) Enhancement and evaluation: How do we enhance text analysis during or after writing (e.g., quality of coherence, style) using corpus linguistic tools? How do we evaluate or compare existing writing assistants (e.g., adequacy, design features, ease of use, lessons learned)?

SUBMISSION INSTRUCTIONS

Please submit your papers via the START/SoftConf submission portal (https://softconf.com/coling2025/AAC-AI25/), following the COLING 2025 templates. Submitted versions must be anonymous and should not exceed 8 pages for long papers and 4 pages for short papers. References do not count toward the page limit, and may be up to 4 pages long. Supplementary material and appendices are also allowed. We also invite papers discussing tools and applications (system demonstrations) related to our workshop topics.

PUBLICATION

All the accepted papers (be it for oral presentation or as poster) will be published as proceedings appearing in the ACL anthology.

PARTICIPATION

The workshop requires a physical presence. If any authors are unable to attend and present in person, alternative arrangements (such as remote presentations or video recordings) may be considered. However, we cannot guarantee these options, as the COLING organizers and local chairs have informed us that they will not provide technical support or online access. Generally, work presented in person will be given preference over work presented virtually.

ORGANIZERS

  • Michael Zock (CNRS, LIS, Aix-Marseille University, Marseille, France)
  • Kentaro Inui (Mohamed bin Zayed University of Artificial Intelligence, UAE; Tohoku University, Japan; RIKEN, Japan)
  • Zheng Yuan (King’s College London and the University of Cambridge, UK)

MORE DETAILS:

Michael ZOCK

Emeritus Research Director CNRS
LIS UMR 7020 (Group TALEP)
Aix Marseille Université
163 avenue de Luminy – case 901
13288 Marseille / France
Mail: michael.zock