Artificial intelligence is a notion first defined in the mid-1950s by John McCarthy. It is the subject of interest in computer science, cognitive science, but also philosophy and social sciences. It has long captivated human imagination with its capabilities, becoming the topic of literature, both popular and scientific.
Artificial intelligence and the assurance of quality in science
The underlying purpose the organizers intended for the event was to raise awareness among the scientific community of the scale and pace of the development of artificial intelligence, of areas where it is rapidly expanding, and of the irreversibility of the processes that the AI phenomenon has been bringing about in many different aspects of life. In the panel sessions, the speakers affiliated with the world of academia and publishing raised the issue of tremendous opportunities, but also of threats inherent in uncontrolled spread of artificial intelligence. They all acknowledged the validity of the general contention: artificial intelligence does indeed affect the quality of publications (not just digital ones), and the way they are prepared. Consequently, it has the capacity to shape the course of research evaluation.
Potential impact of artificial intelligence on the publishing sector
The advantages offered by artificial intelligence are clear and many of its capabilities can be leveraged in the publishing process and the publishing market. Unfortunately, as Małgorzata Adamczak (UAM) pointed out, 'AI can also raise serious concerns.'
The field of academic publishing has many potential applications for artificial intelligence, including identification of market trends, demand forecasting, detection of plagiarism and copyright infringement, translation assistance for certain types of texts, editing and proofreading support, and metadata generation.
Only a portion of this potential depends on AI development. Some of the features have already grown in popularity (academic and legal publishing), while some others are still on trial by publishers. Additionally, AI has the potential to help reduce the environmental impact of the publishing sector by allowing it to more accurately forecast the print run of specific titles.
Artificial intelligence and copyright
One threat that artificial intelligence poses to the publishing industry is the unethical use of certain data in the compilation of resources (databases), known as AI training. The conference attendees insisted on the importance of full disclosure of the works procured for this purpose and their sources. Particularly pertinent is the transparency concerning the extent to which copyrighted works are used. That large language models (including GPT-3, the ChatGPT chatbot engine) are trained on books and other copyrighted texts is something that is now rather well known.
Today, the terms of use of scientific and non-scientific literature in the process of AI training are regulated by the 2019 Directive on copyright and related rights in the digital single market. It ensures that a copyrighted work can only be used for non-commercial purposes, which is why it is critical that artificial intelligence developers cooperate with copyright owners. This is the only way to effectively monitor whether the sources that the artificial intelligence acquires are legal. Indeed, only by being able to review information about datasets and their sources, and consequently being able to exclude certain works, can we ensure that no further copyright infringements occur.
Are works generated by AI under copyright?
Likewise, a considerable number of speakers drew attention to the issue of AI-generated content. The question was raised whether works generated by ChatGPT and other AI systems should be subject of copyright?
According to representatives of scientific publishers, content created by generative AI should be subject to the same copyright eligibility rules as all other content. If the content was created by artificial intelligence and does not reflect the author's (human) creative decisions or personality, the content should not be subject of copyright protection. However, if the author uses artificial intelligence as a tool to create a work that uniquely reflects his or her creativity, it should be protected by copyright law.
AI will go into the publication details
Małgorzata Adamczak expressed the opinion that present-day science is struggling against the scourge of a massive volume of articles and other scientific content that are in fact a product of AI. On the same note, she pointed out the need for readers to be explicitly informed that what they are reading is a text that has in its entirety been generated by artificial intelligence. This would allow them to decide whether they want to read the work through or not. The speaker questioned whether books (and articles) written solely by artificial intelligence should appear on the market. She further noted that should artificial intelligence be used as a tool or for purposes unrelated to content generation per se, such information should also be made disclosed to the reader. Editorial pages, and perhaps even title pages, of future books will probably soon include this new feature.
Digital publications as a valuable component of research evaluation
In considering research evaluation involving digital publications, one cannot help but refer to the place where these publications are produced, i.e., the academic publishing house. Ewa Bluszcz of the University of Lodz Publishing House, stated that publishers should be assessed on a multi-level basis. The speaker listed the following elements of such assessment: a clearly defined mission, operation for the advancement of science, profile of the publishing house, quality of its cooperation with authorship, procedures for qualifying publications for publication, care for compliance with the publishing ethics, and most importantly, the quality of the editorial processes leading to the publication of content in the digital medium. Furthermore, the assessment should also include the techniques of digital collection and sharing of academic content, distribution reach (including open access), and the ways scientific content is promoted.
Tomasz Adamski of the private academic publishing house Exemplum, on the other hand, specifically addressed the criteria for evaluating a digital publication, i.e., communication, visualization, interactivity, identification, sharing, and archiving.
As for the criteria to assess today's academic publishing, they include publishing ethics, peer review, digital accessibility, tags and metadata, digital dissemination and archiving, open access to scientific content.
Artificial intelligence and the issue of academic integrity
Artificial intelligence offers many advantages. It helps optimize our day-to-day email writing, messaging, and marketing presentation processes. With ChatGPT, professionals can now produce articles or conference publications quickly. Yet there is one major concern related to AI models' use: lack of quality control. The risk lies in the fact that the quality of the generated content may fail to meet the standards expected of academic publications, and thereby adversely affect the evaluation of scientific activity. The question of originality of texts prepared based on a database, even the best and most reliably built one, also remains controversial. Therefore, It is vital that appropriate quality control safeguards be put in place to ensure that the content is accurate, relevant, and reliable.
Who, then, is (or shall be) the true creative agent, doer, generator of scientific and non-scientific thought - a human or a machine? The intensity of the processes surrounding artificial intelligence indicates that this question may soon turn into one of the most relevant problems of a transforming world.