Keywords
artificial intelligence, machine learning, open data, open research, Artificial Intelligence and Machine Learning Gateway, sharing
This article is included in the Artificial Intelligence and Machine Learning gateway.
This article is included in the Research on Research, Policy & Culture gateway.
This article is included in the Artificial Intelligence in Academic Research collection.
artificial intelligence, machine learning, open data, open research, Artificial Intelligence and Machine Learning Gateway, sharing
Many large-scale funding bodies have, or are moving towards, new Open Science mandates in their funding requirements, including the European Commission (European Commission, 2022) and the US government (National Institutes of Health, 2023). To meet these requirements, all data utilised in research outputs should conform to the FAIR principles, (Wilkinson et al., 2016), meaning that it should be Findable, Accessible, Interoperable and Reusable. In addition to Open data, to meet Open Science requirements, manuscripts should be published Open Access and underlying software or code utilised in producing the research output should be produced using an open-source option where possible.
For many researchers working in the field of Artificial Intelligence (AI), these concepts are not foreign – for example, the open-source operating system Linux (Wilson et al., 2016) is a popular choice amongst computer scientists, and many share their codes freely on sites such as GitHub, BitBucket or similar. For others, especially those with more industry-focused roles who are often working to achieve patents, Open Science, and in particular data sharing, may be a newer concept. Luckily, there are plenty of resources freely available to help prospective authors with the new requirements of their funding bodies, including F1000’s data policies (Grant, 2022).
The European Commission (European Commission Publications Office, 2018) states that Open Data and data sharing are particularly important for AI and machine learning research due to the large volumes of data required to train machines up to a usable standard. This is particularly notable with respect to the recent rise of AI art, which is trained on art without the permission of the original artists (Ghosh & Fossas, 2022), and may occur in other disciplines as well without either freely shared data to train the AI on and/or modern copyright legislation to protect others’ intellectual property rights.
Unlike many traditional publication models, the F1000Research model is designed specifically for the free sharing of research and its associated materials. It does this in four main ways.
Firstly, through use of Open Access – all articles published on F1000Research are freely available for all at point of publication. By removing the paywalls of traditional publishing, it means that researchers in under-funded countries or institutions also have access to valuable resources and reference work. Additionally, and of particular importance to AI research, news reporters and the general public also have access to this work, giving them the ability to read and understand discussions around AI directly from the source.
Secondly, F1000Research has strict Open Data mandates in line with the current updates to various funding mandates listed above. These mandates require that all data produced or used as part of a research work is available to peer reviewers, readers and researchers interested in reproducing or verifying the research. F1000Research operates under an ‘as open as possible, as closed as necessary’ policy, meaning that all material, including material that can’t be shared (for example, because it contains sensitive patient data) must be declared in the data availability statement along with any reason why it is unable to be shared. More available data is generally excellent news for researchers working in machine learning!
Thirdly, the publication model for F1000Research allows for a huge range of article types across the full research spectrum to be published. This means that valuable studies don’t have to wait until they’re ready to be turned into an original research article to start sharing data – methods articles, data notes and similar are all valuable ways that researchers can share their work prior to the final research article. The model also allows for versioning, which is great for researchers working in AI as it allows for updates to be made to articles as the field evolves, rather than requiring a full new article with only a minor improvement.
Finally, F1000Research employs Open Peer Review. This has several benefits to authors, readers and peer reviewers. For authors, it means that their work can be published online faster as peer review – typically the slowest part in the publication process – occurs after publication. For readers, it means that the full discussion about the article’s merits (and potential oversites or flaws) is freely available to read, allowing them the autonomy to decide for themselves about the article’s quality. For the reviewers themselves, it means they receive full credit for the work they put into this extremely important part of the research publication process, as their review is also published online and can be read and cited.
All these aspects are crucial for fostering an open, sharing research culture, essential to many new funder mandates and incredibly useful for AI researchers who often require a wealth of freely available data for quality machine learning.
Artificial intelligence and machine learning will continue to be at the forefront of the development of new computer-based technologies for the foreseeable future. However, with the issues arising from unconsidered applications, like students using ChatGPT to write homework essays (Cotton et al., 2023) and the potential for a rise in AI-generated research articles with very little oversight (Dergaa et al., 2023), it is important to share as much research and data openly as possible. This allows users of the technology to fully understand the work and data behind the technology, allows legal experts and policymakers across both governments and industry to ensure their guidance is current and relevant, and it allows other AI researchers to make further developments and decisions based on a thorough understanding of what has come before. Ethical considerations and discussion, as well as topics on any of the above are welcome in the Artificial Intelligence and Machine Learning Gateway.
The F1000Research model is particularly well-suited to facilitate the open, rapid dissemination of research works related to AI and related fields, and its applications.
Firstly, following the high-level policy decisions made by many funders already, it is essential to have a publication venue that already complies with their new requirements, and has done for many years (Pencheva et al., 2018). This ensures that publications meeting these requirements are handled smoothly and expertly by F1000Research’s internal editors, who will guide authors through ensuring they comply with their funding mandates.
Secondly, trustworthy data for AI is of utmost importance to ensure accurate and efficient machine learning strategies (Liang et al., 2022). While outliers or ‘untrustworthy’ data is less important in human education, due to our ability to recognize oddities or issues with datasets, AI takes all data at face value unless taught otherwise. Therefore, clean data is especially important when creating new AI software for new applications. Fostering the sharing of data that has already been analyzed and understood, and potentially used to train AI, is a cornerstone of the F1000Research model and why this Gateway has the potential to be a go-to resource for AI scholars and industry professionals alike.
Finally, most current publication venues are limited in publications to technical or applied research and excluding relevant contributions from the humanities and social sciences. The Artificial Intelligence and Machine Learning Gateway has the scope and interdisciplinary nature required to foster important sociological and other humanities research as well, providing a space where important discussions can occur about the ethics, legality and social aspects of new AI-powered technologies.
Therefore, the Artificial Intelligence and Machine Learning Gateway will provide an Open Science space for interdisciplinary publications on all aspects of AI, bridging the many different communities to create open and thoughtful discussions. It can provide a useful resource and collaborative space to guide researchers, authors and industry professionals through this new era of artificial intelligence.
Views | Downloads | |
---|---|---|
F1000Research | - | - |
PubMed Central
Data from PMC are received and updated monthly.
|
- | - |
This article is an Editorial and has not been subject to external peer review.
Provide sufficient details of any financial or non-financial competing interests to enable users to assess whether your comments might lead a reasonable person to question your impartiality. Consider the following examples, but note that this is not an exhaustive list:
Sign up for content alerts and receive a weekly or monthly email with all newly published articles
Already registered? Sign in
The email address should be the one you originally registered with F1000.
You registered with F1000 via Google, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Google account password, please click here.
You registered with F1000 via Facebook, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Facebook account password, please click here.
If your email address is registered with us, we will email you instructions to reset your password.
If you think you should have received this email but it has not arrived, please check your spam filters and/or contact for further assistance.
Comments on this article Comments (0)