Large Language Models in Academic Publishing

Introduction

The 'traditional' academic publishing process comprises four key steps: (i) manuscript preparation that involves the conduct of original research, synthesis of the literature, and drafting of a paper, report, or thesis to present the findings of the research; (ii) peer review where the manuscript is evaluated by experts in the area in order to provide the author with feedback and insights on how to improve upon the work and manuscript, (iii) editing to ensure that language, formatting, citation, and referencing styles meet the publisher's requirements, and (iv) the production, marketing and distribution steps where the finished work is availed in either print or online formats for consumption by its target market (Lambert Academic Publishing, 2023).

These steps, in the traditional sense, have been human-intensive but advances in Information and Communication Technologies (ICTs) have served to enhance critical aspects of the publication process, such as data collection, transcription and analysis, manuscript preparation, formatting and layout, citation and reference management, and distribution by way of electronic or online publishing and marketing (Harman, 2023). In recent times, the role of ICTs in publication has progressed beyond simply managing processes and documents to performing additional tasks that have always required human input. There are now a significant number of software systems that are capable of tasks such as writing, reviewing, and editing manuscripts. These capabilities emanate from a branch of computing called Artificial Intelligence (AI).

Artificial Intelligence is a subfield of computing that examines the development and use of computer systems that can perform tasks ordinarily performed by humans. These include (i) perception of the physical world through vision, sound, smell, touch, and taste, (ii) manipulation of the environment through movement, (iii) reasoning through situations in order to arrive at optimal paths of action, (iv) communicating through receiving of audio or textual input, processing it and formulating a suitable response, and (v) learning through experience in order to improve performance of tasks over time (Williams, 1983). AI has found application in diverse domains such as agriculture, commerce, entertainment, robotics, human resource management, healthcare, the motor vehicle industry, social media, marketing, finance, security, and education (Simplilearn, 2022).

Regarding capabilities, AI systems can be generally categorized as (i) Artificial Narrow Intelligence systems capable of completing concrete actions and cannot learn independently. Examples include spam filters, search engines, virtual assistance, industrial robots, recommendation systems, fraud detection systems, machine translation, medical diagnosis systems, natural language processing systems, and self-driving cars. (ii) Artificial General Intelligence systems that can learn, think, and perform at an almost human level, and (iii) Artificial Super Intelligence systems that can surpass the knowledge and capabilities of human beings (Joshi, 2019).

AI-powered systems continue to grow in capability, adoption, and acceptance in various domains. Recently, they have found practical applications in the creative domain and can generate new text, image, and audio content at a level close to that of human beings. This class of AI systems is called Generative Artificial Intelligence (GenAI) systems. They learn and reproduce the patterns and structure in pre-existing data to create new content (Lawton, 2023). In academic publishing, GenAI has found application in writing, reviewing, editing, and marketing manuscripts. This chapter examines their application, benefits, and challenges in the context of academic publishing and further makes recommendations for their beneficial and responsible use in practice.