In July, HEPI, with support from publishers Taylor & Francis, held a roundtable dinner to discuss AI use to advance the use and impact. This blog explores some of the topics that emerged from the discussion.
In the near future, if you travel to major train stations, you may see a video of someone using British sign language along with a board that gives the train time. This is an AI-generated signatories, turning the often difficult measurement station announcements into sign language so that the hearing impaired can understand what is being said. This is just an example of how artificial intelligence is used in the real world.
The question that this Roundtable focused on was how AI could be used to advance translation research. In other words, it enables research that is guided by curiosity and turns it into real-world applications. What role do academic leaders and publishers play in shaping the ethical, comprehensive and innovative use of AI in such research? How can AI enhance interdisciplinary collaboration? Also, what are the potential barriers, ethical dilemmas and risks that come with the process?
Discussions attended by senior universities and research leaders, publishers and funders will be held under Chatham House rules, where speakers will express their opinions on understandings that they become non-subjects.
Benefits and risks
The speaker agreed that AI has great potential to enable researchers to analyze large datasets cheaply, quickly and accurately. They noted that AI provides a simple language overview of research and can be useful in presenting it in a variety of formats, including multilingual and multimedia content, while also opening up useful ways for learning societies to spread research findings among member practitioners.
However, risks have also been identified. How does AI use affect creativity and critical thinking among researchers? How can scholars protect bias and ensure transparency in the data underlying AI tools? And what about environmental concerns? What about maintaining an AI system that attacks energy and managing e-waste? Most worrying, if AI is involved in research and its applications, if something goes wrong, who will ultimately be accountable?
Such concerns were addressed in a guide for researchers who embrace AI in good faith, published by Research Integrity Office Ukrio in June. https://ukrio.org/wp-content/uploads/embracing-ai-with-integrity.pdf.
Representatives from the Round Table were told that one of the messages to draw from this guide was that researchers using AI should ask three important questions.
- Who owns the information entered into AI?
- Once you reach AI, who owns the information?
- Who owns the output?
Working together
Collaboration is important, one speaker said. It means dismantling existing academic silos and inviting them to experts responsible for applying AI-driven research. It is also important to consider the wider picture and the type of society we want to be in the future.
One concern identified by the Round Table was that the power on AI systems is concentrated in the hands of a small number of people. This means that they are creating division in terms of access to information and resources rather than addressing social issues.
“We're not in the age of AI that we actually want,” one speaker said. “We are in an age of AI given to us by big technology.”
Addressing this issue could involve the development of new regulatory and legal frameworks, particularly to establish accountability. Practitioners are particularly concerned about “where the bag stops” and how, for example, potentially transformative AI diagnostic tools can be used in a safe way.
Others on the Round Table were concerned that putting researchers a large part of their ethical responsibility for AI might discourage boundary testing.
“When you do research, you will never be able to fully control that, or you will never do anything novel.” Therefore, responsibility must be shared among researchers, implementers and users. This means that everyone needs education in AI, so they understand the tools they are given and how to use them effectively.
Reliable data
Being able to rely on the underlying datasets used in AI is essential, said one speaker who welcomed the government's decision to open public datasets through the AI Opportunity Action Plan https://www.gov.uk/government/publications/ai-opportunities-action-plan/ai-opportunities-portunities-action-partunity-pran and curate-plan and curate-curate-fublication.
There was a difference between research driven by commercially available AI tools when they were unable to see “inside the black box” and research based on AI tools where datasets and algorithms are reliable and transparent. The former was like presenting a research paper that provided introductions, results and analysis without explaining the methodology. That was proposed.
User education
However, AI is not only about the data it is based on, but also about the capabilities of the people who use it. How can higher education institutions ensure that students and researchers, especially early career researchers, have the know-how that requires proper use of AI? ( Taylor and Francis AI Policy It might be interesting here. )
In order to publish recommendations later this year, it was noted that an independent review of curriculum and rating systems in UK schools is likely to be a missed opportunity when it comes to ensuring students will be enrolled in college with AI skills.
Meanwhile, politicians often lack expertise in this field and have struggled to establish a suitable framework for AI research.
This is a problem as the field moves so fast. It suggested that rather than waiting for action from policymakers and regulatory frameworks, researchers should either use AI or take the risk of the UK being left behind.
Social vision
The Round Table agreed that it is not just the responsibility of the academia to decide all of this. However, what academic research could be useful was to bridge the gap in AI development that large commercial companies ignored because they prioritized their business models.
Here, researchers, including the arts and humanities, may be important in determining what society hopes to achieve in the end for AI. Otherwise, one speaker suggested that it would be driven by the “art of possibilities.”
Meanwhile, what skills do universities want researchers to have? Some raised fear that outsourcing work to AI could mean that researchers are being desked. Evidence already suggests that AI use can reduce students' metacognition, that is, their understanding of their own thought processes.
“If it's important for researchers to be able to translate their findings, don't let the machine do that,” said one speaker. Another questioned whether researchers should use tools that they could not understand.
Artificial colleagues
One suggestion was that researchers should use it to enhance existing practices rather than outsource research to AI.
While some were concerned about the impact of AI on creativity, one speaker suggested that by tuning AI tools to explore concepts at the edge of scientific consensus, it could be used to wrap more original approaches than humans would achieve on their own.
Another positive identification is that AI bias can be a problem, but it can be more easily identified than human bias.
I heard that the Round Table should be about teamwork. AI is considered another co-worker – you can't do all the work there.
“AI detects your bias, it reviews your work, it supports that process, but you shouldn't think,” a message from one speaker. “In the end, it should come back to humanity.'
Taylor & Francis are HEPI partners. Taylor & Francis supports a diverse community of experts, researchers and knowledge makers around the world to accelerate and maximize the impact of their work. We are leaders in our field, open to all fields and have one of the largest portfolios of humanities and social sciences. Building on a heritage of more than 200 years of academic publication, our expertise advances trustworthy knowledge that drives human progress. Under the imprint of Taylor & Francis, Routledge and F1000, it publishes 2,700 journals and 8,000 new books each year, and partners with over 700 academic societies.
We will work together to develop HEPI policy notes on the use of AI in advances in translation research. If your institution has great case studies or AI-related translation approaches, we'd love to hear from you. Please tell us more about your job. Please contact us by email [email protected].
