Friday, December 20, 2024

 

Lost score revives sound of music and voices from centuries past

A fragment of ‘lost’ music found in the pages of Scotland’s first full-length printed book is providing clues to what music sounded like five centuries ago.

Peer-Reviewed Publication

University of Edinburgh

Aberdeen Breviary : Volume 1 or 'Pars Hiemalis' 

image: 

Fragment of music from the Aberdeen Breviary : Volume 1 or 'Pars Hiemalis'. Image courtesy of the National Library of Scotland

view more 

Credit: Images are courtesy of: National Library of Scotland and may be used under the Creative Commons Attribution 4.0 International Licence. Caption - Aberdeen Breviary : Volume 1 or 'Pars Hiemalis'. Link to downloads and additional credit information - https://digital.nls.uk/aberdeen-breviary/archive/102878258?mode=fullsize.

Lost score revives sound of music and voices from centuries past

A fragment of ‘lost’ music found in the pages of Scotland’s first full-length printed book is providing clues to what music sounded like five centuries ago.

Scholars from Edinburgh College of Art and KU Leuven in Belgium have been investigating the origins of the musical score – which contains only 55 notes – to cast new light on music from pre-Reformation Scotland in the early sixteenth-century.

Researchers say the tantalising discovery is a rare example of music from Scottish religious institutions 500 years ago, and is the only piece which survives from the northeast of Scotland from this period.

The scholars made the discovery in a copy of The Aberdeen Breviary of 1510, a collection of prayers, hymns, psalms and readings used for daily worship in Scotland, including detailed writings on the lives of Scottish saints. Known as the ‘Glamis copy’ as it was formerly held in Glamis Castle in Angus, it is now in the National Library of Scotland in Edinburgh.

Despite the musical score having no text, title or attribution, researchers have identified it as a unique musical harmonisation of Cultor Dei, a night-time hymn sung during the season of Lent.

The Aberdeen Breviary came from an initiative by King James IV who issued a Royal Patent to print books containing orders of service in accordance with Scottish religious practices, rather than needing to rely on importing texts from England or Europe.

The researchers say the composition is from the Aberdeenshire region, with probable links to St Mary’s Chapel, Rattray – in Scotland’s far northeastern corner – and to Aberdeen Cathedral.

The discovery was made as researchers examined numerous handwritten annotations in the margins of the Glamis copy.

Of primary interest to the scholars was a fragment of music – spread over two lines, the second of which is approximately half the length of the first – on a blank page in a section of the book dedicated to Matins, an early morning service.

The presence of the music was a puzzle for the team. It was not part of the original printed book, yet it was written on a page bound into structure of the book, not slipped in at a later date, which suggests that the writer wanted to keep the music and the book together.

In the absence of any textual annotations on the page it was not clear whether the music was sacred, secular or even for voices at all, the researchers say.

After investigation they deduced it was polyphonic – when two or more lines of independent melody are sung or played at the same time. Sources from the time say this technique was common in Scottish religious institutions, however very few examples have survived to the present day.  

Looking closer, one of the team members realised that the music was a perfect fit with a Gregorian chant melodyspecifically that it was the tenor part from a faburden, a three- or four-voice musical harmonization, on the hymn Cultor Dei.

David Coney, of Edinburgh College of Art, who discovered the identity of the music, said: “Identifying a piece of music is a real ‘Eureka’ moment for musicologists. Better still, the fact that our tenor part is a harmony to a well-known melody means we can reconstruct the other missing parts. As a result, from just one line of music scrawled on a blank page, we can hear a hymn that had lain silent for nearly five centuries, a small but precious artefact of Scotland’s musical and religious traditions.”

As well as uncovering lost sounds within its pages, researchers have also traced how the Aberdeen Breviary may have been used, and by whom, over its long history. At one time used as the private service-book of the illegitimate son of a high-ranking chaplain at Aberdeen Cathedral, himself a rural priest, it would later become a treasured family heirloom of a Scottish Catholic whose travels led him from post-Reformation Scotland to the capitals of the Habsburg and Ottoman Empires.

Lead author, Dr Paul Newton-Jackson, of KU Leuven, said: “The conclusions we have been able to draw from this fragment underscore the crucial role of marginalia as a source of new insights into musical culture where little notated material survived. It may well be that further discoveries, musical or otherwise, still lie in wait in the blank pages and margins of other sixteenth century printed books held in Scotland’s libraries and archives.”

In 2023, Dr Newton-Jackson was also a post-doctoral fellow at the University of Edinburgh’s Institute for Advanced Studies in the Humanities.

Dr James Cook, of Edinburgh College of Art, said: “For a long time, it was thought that pre-Reformation Scotland was a barren wasteland when it comes to sacred music. Our work demonstrates that, despite the upheavals of the Reformation which destroyed much of the more obvious evidence of it, there was a strong tradition of high-quality music-making in Scotland’s cathedrals, churches and chapels, just as anywhere else in Europe.”

The study is published in the Journal Music and Letters.

Music from 16th Century Scotland - Cultor Dei Memento 

 

Tracking other people’s minds in communication


Max Planck Institute for Psycholinguistics




Language and social cognition are fundamental to human communication. But how do these capacities interact? In a review paper published in Trends in Cognitive Sciences, researchers from the Max Planck Institute of Psycholinguistics (MPI) in Nijmegen and Yale University show how language and social cognition are integrated in real time. The authors propose a new ‘mind-tracking’ model of communication, in which social micro-processes play a fundamental role in language production and comprehension.

 

To communicate successfully, people need to track other people’s minds. For example, when your host says ‘it’s getting dark outside’, you may infer that she wants you to leave. Social cognition—the capacity to understand other people’s beliefs, desires or intentions—is vital for using and understanding such non-literal language. Traditionally, the connection between language and social cognition was thought to occur mainly at the sentence level.

While these global processes are fundamental to human communication, local processes at the word level are equally fundamental, in our view”, says MPI’s Paula Rubio-Fernández, senior investigator and co-author of the study. “Communication is full of social micro-processes that happen in both language production and language comprehension, which recruit social cognition in real timeThe interdependence between language and social cognition in human communication is deeper and more pervasive than originally thought”.

Everyday examples of social micro-processes are choosing definite or indefinite articles depending on whether we are talking about something familiar or new to the listener (‘We bought the house’ vs ‘We bought a house’) or choosing demonstratives (‘this cup’ or ‘that cup’) to guide the listener’s attention to the intended referent.

Recent advances in computational models of social cognition, including our own, offer support for this view, allowing us to model the primary cognitive processes that we represent in other minds. Our findings broaden the scope of the relationship between language and social cognition, relative to traditional accounts”, Rubio-Fernández concludesWe plan to further this line of work by investigating referential communication in multimodal, naturalistic interaction, focusing not only on speech and sign, but also on gaze and gesture.”

 

Publication

Paula Rubio-Fernández, Marlene D. Berke & Julian Jara-Ettinger (2024). Tracking Minds in Communication. Trends in Cognitive Sciences.

 USE VALUE VS EXCHANGE VALUE

Personalized gifts create lasting emotional connections and enhance self-esteem – new research



Study identifies emotional benefits of both giving and receiving personalised gifts



University of Bath



There is joy in giving but research from the University of Bath shows that  personalisation can transform a gift from a mere object into a more meaningful experience that raises recipients’ self-esteem and makes them feel more cherished.

Researchers in the UK, France and Switzerland demonstrated that personalised gift giving invoked a unique emotional response, generating what they dubbed ‘vicarious pride’ on the part of the recipient.

“Gift-giving is an age-old tradition, but in today’s world, personalisation has become a powerful way to make gifts stand out. Our research paper ‘You designed that yourself for me? Vicarious pride in customized gift exchange’ explores why customised gifts are so appreciated by recipients,” said Dr Diletta Acuti, marketing expert at the University of Bath School of Management.

“Vicarious pride is a sense of satisfaction that mirrors what the giver feels after having put thought and effort into creating something unique. Imagine your friend personalising a chocolate bar with your favourite flavours on the wrapper or writing your name on a leather journal. You don’t just appreciate the care and intention they put into crafting that gift; you feel them,” she said.

Researchers from the University of Bath, emlyon and SKEMA business schools in France, and Universita della Svizzera italiana (USI) in Lugano conducted four experimental studies.

In the first, 74 participants gifted clothes to a friend – some personalised. Gift appreciation was assessed by asking recipients to change any items of the gift they did not like. Recipients of personalised gifts changed fewer outfit items, reflecting greater appreciation.

In the second study, researchers showed 134 participants videos of T-shirt‐selection processes – one showing gift customisation, the other a gift selection made by surfing different websites. The study found appreciation was higher in the case of the customised gift, regardless of the time and effort the giver put into the process.

Studies 3 and 4, conducted online with a mug and wristwatch as gifts, confirmed that customisation increased appreciation and enhanced recipients’ self-esteem.

“The results showed that, when choosing a gift this festive season, personalisation can be a game-changer. But it’s not just about selecting a customisable option: you also need to communicate that effort to your recipient,” Dr Acuti said.

“Sharing why you chose elements of the gift or the thought that went into it will make the recipient appreciate it even more. Indeed, this additional effort helps them to connect with the pride you felt in your choices, making the gift even more meaningful,” she said.

The study noted that personalisation also had implications for sustainability, as recipients were more like to take care of gifts that they valued more. Brands could benefit too by helping buyers to highlight their personal touch.

“Using ‘made by’ signals – such as including the giver’s name, a short message about the process or a visual representation of the customisation – can make things even more impactful. These small additions reinforce the emotional connection between the giver and the recipient,” Dr Acuti said.

ENDS/tr

Notes to editors

  • For more information contact the University of Bath Press office at press@bath.ac.uk

The University of Bath 

The University of Bath is one of the UK’s leading universities, with a reputation for high-impact research, excellence in education, student experience and graduate prospects.  

We are ranked in the top 10 of all of the UK’s major university guides. We are also ranked among the world’s top 10% of universities, placing 150th in the QS World University Rankings 2025. Bath was rated in the world’s top 10 universities for sport in the QS World University Rankings by Subject 2024.

Research from Bath is helping to change the world for the better. Across the University’s three Faculties and School of Management, our research is making an impact in society, leading to low-carbon living, positive digital futures, and improved health and wellbeing. Find out all about our Research with Impact: https://www.bath.ac.uk/campaigns/research-with-impact/

Paris Declaration calls for data-driven forensics to spearhead the fight against fake science



Research integrity champions say Forensic Scientometrics (FoSci) will decontaminate “polluted” science and scholarly literature



Digital Science

Forensic Scientometrics (FoSci) Paris Declaration - quote from Dr Leslie McIntosh 

image: 

"What FoSci does is shine a light on questionable or deceitful practices in the world of science," says Dr Leslie McIntosh, VP of Research Integrity, Digital Science.

view more 

Credit: Digital Science.




Supporters of research integrity have signed a new declaration calling for data-driven forensics – known as Forensic Scientometrics (FoSci) – to lead the charge in detecting, exposing and even preventing fake science.

The Forensic Scientometrics (FoSci) Paris Declaration was drafted during an event in Paris last week organized and run by Digital Science’s VP of Research Integrity, Dr Leslie McIntosh. The event was hosted at Institut Universitaire de France (IUF) by Dr Guillaume Cabanac, research integrity “sleuth” and professor at the University of Toulouse, as part of his research chair titled “Decontamination of the scientific literature.”

The event involved researchers, experts, and professionals from around the world who are committed to upholding research integrity, many well-known sleuths among them. Attendees signed the declaration over the following weekend.

As the Declaration states, “Trustworthy science risks being obscured by a small but growing corpus of papers, people, organizations, and potentially governments polluting the integrity of research.”

And: “We care deeply about science, and we believe firmly in the ability of scientific study to decontaminate the scholarly literature. As a collective, we intend to do whatever we can to promote the consistency and reliability of scientific research output.”

“We want to dispel this pollution by flagging problematic papers, actors, and systems, mitigating the effects and disincentivizing such behavior in the future. Our goals are to prevent these errors from spreading, to promote better policies for scientific endeavours, and to safeguard the positive impact of science on society.”

FoSci is a forensic, data-driven initiative to uphold scientific integrity and public trust in science. It combines forensic investigation and scientometrics, which is the study of how research is shared and built upon. FoSci uncovers patterns that uphold or threaten the integrity of science itself.

The problems currently researched by forensic scientometricians include: author misrepresentation, data manipulation, fake conferences, image duplication, misconduct (including fabrication, falsification, and plagiarism), papermill operations, questionable research practices, sale of authorship and citations, sneaked references, stealth corrections, and tortured phrases.

The Declaration states that these problems have widespread and potentially damaging implications, through the citation of fraudulent research in patents, clinical guidelines, government policy, and more.

Dr McIntosh, one of the co-founders of the FoSci movement, said: “Forensic scientometrics is needed now more than ever. Scientific achievement is critical to our society’s health and wellbeing, to our economic and social prosperity, but we also live in a time when the community’s trust in science is constantly being eroded.

“What FoSci does is shine a light on questionable or deceitful practices in the world of science. Through collective action, we want to motivate those involved in producing and disseminating scientific research to produce consistent, valid, and high-quality work.”

Dr Cabanac said: “Our gathering of institutions, journalists, publishers, sleuths, and a leading scientometric data provider proved highly stimulating and productive. Meeting in person created a synergy that we, as a community, plan to sustain and put at the service of science.

“Unreliable bricks weaken the wall of knowledge that researchers have been building for centuries, one publication at a time. Collective action is required, both curative to prevent humans and AIs from learning from these, and preventive to design methods to stop new forms of misleading contents from entering the scientific record. This declaration is a call for action: join us.”

The FoSci Paris Declaration has made the following key commitments:

  • Advocate for transformation
    • Open a dialogue with policymakers to design de-incentivizing strategies to tackle the mass production of problematic papers
    • Advocate for reform of institutions involved in scientific research based on the group’s findings
       
  • Develop expertise and share knowledge
    • Facilitate training for researchers and professionals exploring these questions 
    • Share and provide research and data in the FoSci community
    • Establish a regular cycle of professional meetings
    • Improve the tools and methods of forensic scientometrics
       
  • Improve the group’s ability to communicate its findings
    • Inform editorial boards, publishers, research institutions, governments and all relevant involved parties about the group’s work 
    • Participate in building software and tools to enable the reproducibility of their forensics findings
    • Establish points of contact between FoSci members and concerned organizations

The Paris event and its declaration are the culmination of a year of awareness-raising activities for Dr McIntosh, who has held workshops on FoSci in Athens, Los Angeles, Darwin and Sydney throughout 2024.

The signatories to the Paris Declaration hope that FoSci will become internationally recognized and taught at research institutions, particularly within research administration teams, but also among the academic community. 

See the full Forensic Scientometrics (FoSci) Paris Declaration here

Find out more about Forensic Scientometrics

 

About Digital Science

Digital Science is an AI-focused technology company providing innovative solutions to complex challenges faced by researchers, universities, funders, industry and publishers. We work in partnership to advance global research for the benefit of society. Through our brands – Altmetric, Dimensions, Figshare, IFI CLAIMS Patent Services, metaphacts, OntoChem, Overleaf, ReadCube, Scismic, Symplectic, and Writefull – we believe when we solve problems together, we drive progress for all. Visit digital-science.com and follow Digital Science on Bluesky, on X or on LinkedIn.


Media contact

David Ellis, Press, PR & Social Manager, Digital Science: Mobile +61 447 783 023, d.ellis@digital-science.com

 

New computational method uncovers surprising variability in Neolithic building practices



The Hebrew University of Jerusalem
Aerial photograph of structures at the Natufian site of Nahal Ein Gev II, demonstrating the variety of shapes. 

image: 

Aerial photograph of structures at the Natufian site of Nahal Ein Gev II, demonstrating the variety of shapes.

view more 

Credit: (Credit - Naftali Hilger)




This study reveals that the early architectural development in the Near East was far more complex than previously thought, challenging the traditional narrative of a straightforward transition from round to rectangular structures during the Neolithic period. The research introduces a novel computational approach to analyze architectural remains, providing an objective and repeatable method for examining building outlines. This methodology uncovers insights that qualitative methods might overlook, including quantification of defined angles in structures as early as the Natufian period, suggesting advanced building techniques were in use earlier than believed. The study also highlights the architectural diversity during the Natufian and early Neolithic period, suggesting that building codes and more standard practices had not yet emerged. These findings provide new insights into the social and technological changes accompanying the rise of permanent settlements and agricultural economies.

Link to Photos: https://drive.google.com/drive/folders/1IccdctsJHpAiAPk7VV9REsl4I5jQHL81?usp=sharing 

A recent study led by researchers from the Institute of Archaeology at the Hebrew University—Hadas Goldgeier, Dr. Antoine Muller, and Prof. Leore Grosman—introduces a new computational method to analyze the architectural development of early settlements. By offering an objective and repeatable approach to examining architectural remains, the study gives a new perspective on the transition from round to rectangular structures during the Neolithic period in the Near East.

Architectural forms have long been tied to societal shifts such as changes in social organization, demographics, and economic strategies. However, traditional interpretations of early architecture have relied on qualitative methods, which may oversimplify complex trends. This study provides a fresh perspective by digitizing building outlines and employing computational tools to objectively quantify their two-dimensional morphology. Using measures like the directionality of normal vectors and minimum angles, the team analyzed 118 structures from 23 sites across the Mediterranean region and Jordan Valley, dating from the Natufian culture to the early Neolithic (15,000–8,500 years ago).

The results revealed a far more nuanced picture than the conventional “round-to-rectangular” narrative suggests. Structures with right angles, previously associated with later architectural phases, were found as early as the Natufian period, reflecting significant advances in building technology at an earlier stage than previously thought. Moreover, the variability in structure shapes was highest during the Natufian and early Neolithic periods, reflecting a lack of formalized conventions. Later in the Neolithic, however, architectural forms showed greater uniformity, potentially signaling the emergence of codified building traditions.

This study not only challenges traditional interpretations but also highlights the social and technological changes accompanying the transition to permanent settlements and agricultural economies. The computational methodology offers an innovative lens to understand architectural history, demonstrating how quantitative analyses can uncover trends that remain hidden in qualitative approaches.

By emphasizing the dynamic variability and technological ingenuity of early builders, this research sheds new light on the inception of architecture in the Levant and underscores the importance of integrating modern analytical tools into archaeological studies.


An aerial photograph of the Natufian site of Nahal Ein Gev II.

Credit

(Credit - Naftali Hilger)


An aerial photograph of the Natufian site of Nahal Ein Gev II.

Credit

(Credit - Naftali Hilger)


 

Machine psychology – a bridge to general AI




Linköping University




Artificial intelligence that is as intelligent as humans may become possible thanks to psychological learning models, combined with certain types of AI. This is the conclusion of Robert Johansson, who in his dissertation from Linköping University has developed the concept of Machine Psychology and how it can contribute to AI development.

Artificial General Intelligence (AGI) has been the holy grail of AI research since the 1950s. So far, humanity has not managed to create an artificial intelligence that can solve intellectual tasks in the same way as humans. But some scientists believe that this may happen as soon as within the next five years. 

One of them is Robert Johansson, who recently defended his PhD dissertation in computer science at Linköping University. But unlike the gloomy scenarios of the future that are often conjured up about AGI in popular culture, he believes that it can be of benefit to humanity.

“Yes, I’m sure of that! It changes everything we do and isn’t just a continuation of ordinary AI – it’s something completely different. I think AGI will have a broad impact in society in a completely different way than today’s AI. You can create a new type of agent, such as virtual researchers or psychologists – but also a lot more,” says Robert Johansson.

At the same time, he sees challenges with technology developments in a troubled world. AGI could be used for various purposes, for example, to create division in society.

“It’s a technology we really have to handle with care. On the other hand, I also think that artificial general intelligence can help counter several destructive developments in society. It might help us human beings to emerge as more loving. I’m open to AGI being able to help us develop in that way as well,” says Robert Johansson.

But developing an artificial intelligence at the same level as a human being is a monumental challenge. Different researchers try to approach the problem in different ways. Some believe that large-scale language models like ChatGPT are the way to go, while others suggest simulating the brain. The path that Robert Johansson has chosen to follow can be called principle-based. This means that he has tried to identify important psychological learning principles that could explain intelligence and then implement them in a computer.

In parallel with his doctoral studies at LiU, he works as a teacher and researcher in clinical psychology at Stockholm University, where he is also an associate professor, a background that he used in his thesis Empirical Studies in Machine Psychology. 

“I’ve used the principles of modern learning psychology to approach the issue of learning, thinking and intelligence. Then I used a specific form of adaptive artificial intelligence which is a logic system where I try to implement learning psychology,” says Robert Johansson, who is now getting his second PhD. 

The logic system is called the Non-Axiomatic Reasoning System, NARS, and is designed to operate without complete data, with limited computational power and in real time. It provides a flexibility that is important for dealing with problems that may arise in the real world.

The combination of NARS and learning psychology principles constitutes an interdisciplinary approach that Robert Johansson calls Machine Psychology, a concept he was the first to coin but more actors have now started to use, including Google DeepMind.

The idea is that artificial intelligence should learn from different experiences during its lifetime and then apply what it has learned to many different situations, just as humans begin to do as early as the age of 18 months – something no other animal can do.

“If you manage to implement this in a computer, then you’ve really cracked the riddle of artificial intelligence on a human level. And I think psychology will be a potentially crucial science for AGI,” says Robert Johansson.

Whether humanity will succeed in creating its intellectual equal within five years remains to be seen. But then there are a lot of other aspects that need to be taken into account beforehand, according to Robert Johansson. 

“We live in a society with laws and rules and moral premises. It’s necessary to take a stance on how to view the rights and obligations of such agents. Maybe AGI will just be a program you run in your browser, but I think that if it has a consciousness it’s still some kind of life.”

 

AI system unveils hidden dark drug jargon unknown even to police officers



The University of Electro-Communications
Detection of Compound-Type Dark Jargons Using Similar Words 

image: 

A groundbreaking AI-driven method has been developed to identify drug trafficking-related dark jargons on social media platforms like X (formerly Twitter).

view more 

Credit: Dr. Yuichi Sei




Drug trafficking and other criminal activities on social media have become a growing social concern. To evade detection by law enforcement and automated monitoring systems, offenders use dark jargons, often combining multiple common words to form "dark jargon" - covert dark jargons that are difficult to identify. Once authorities recognize these dark jargons, offenders quickly switch to new ones, making it a constant race to keep up with the evolving words.

 

A team of researchers from the University of Electro-Communications (UEC), including Takuro Hada, Yuichi Sei, Yasuyuki Tahara, and Akihiko Ohsuga, has developed a groundbreaking method to tackle this challenge. Their new system leverages the power of artificial intelligence to detect these elusive, compound-type dark jargons with greater accuracy than ever before. This innovative approach focuses on the relationships between words within posts to identify terms that are often used together as part of a dark jargon.

 

Traditional methods struggle to detect dark jargons when it consists of multiple words because most text analysis tools automatically break down sentences into smaller word units. This process often separates words that should be treated as a single phrase, like "Green-Crack" or "Pineapple-Chunk," both known as dark jargons for illegal substances. The new AI system overcomes this by identifying pairs of words that frequently appear together in similar contexts. By analyzing large datasets of social media posts, the AI can recognize when two or more words form dark jargons, even if it's a newly emerging term.

 

The impact of this development is significant. In experiments, the system identified more dark jargons than previous methods, with a 7% improvement in accuracy. Notably, during interviews with police officers experienced in organized crime investigations, 93% of the newly detected dark jargons were confirmed as previously unknown. This highlights the potential of the AI to reveal emerging dark jargons that evade current detection efforts.

 

With crime increasingly moving to online spaces, this technology provides law enforcement with a vital tool to stay ahead of offenders. By automating the detection of new dark jargons, police forces and monitoring systems can respond more swiftly, reducing the risk of illegal activity on social media platforms. This innovation not only strengthens public safety but also supports social media companies in keeping their platforms safer for all users.

 

This research marks a major step forward in the fight against online crime, offering an adaptable, AI-driven solution that evolves as quickly as the dark jargon it aims to detect. With its high accuracy and ability to detect previously unknown terms, this system has the potential to become a key element of future cybercrime prevention strategies.

 

In Japan, where 'yami baito' - a term referring to illegal part-time jobs often linked to criminal activities - has become a growing social issue, this AI technology is expected to play a critical role. Detecting the dark jargon used in online recruitment for these illicit activities, supports law enforcement and enhances public safety.

ChatGPT errors show it cannot replace finance professionals, yet



Washington State University




PULLMAN, Wash. – While large language models like ChatGPT can do well when choosing multiple-choice answers on financial licensing exams, they falter when dealing with more nuanced tasks.

A Washington State University-led study analyzed more than 10,000 responses to financial exam questions by the artificial intelligence language models BARD, Llama and ChatGPT.

The researchers asked the models to not only choose answers but also explain the reasoning behind them, then compared those text answers to those by human professionals. While two versions of ChatGPT performed the best at these tasks, they still showed a high level of inaccuracy with more advanced topics.

“It’s far too early to be worried about ChatGPT taking finance jobs completely,” said study author DJ Fairhurst of WSU’s Carson College of Business. “For broad concepts where there have been good explanations on the internet for a long time, ChatGPT can do a very good job at synthesizing those concepts. If it’s a specific, idiosyncratic issue, it’s really going to struggle.”

For this study, published in the Financial Analysts Journal, Fairhurst and co-author Daniel Greene of Clemson University used questions from licensing exams including the Securities Industry Essentials exam as well as the Series 6, 7, 65 and 66.

To move beyond the AI models’ ability to simply pick the right answer, the researchers asked the models to provide written explanations. They also chose questions based on specific job tasks financial professionals might actually perform.

“Passing certification exams is not enough. We really need to dig deeper to get to what these models can really do,” said Fairhurst.

Of all the models, the paid version of ChatGPT, version 4.0, performed the best, providing answers that were the most similar to human experts. Its accuracy was also 18 to 28 percentage points higher than the other models. However, this changed when the researchers fine-tuned the earlier, free version of ChatGPT 3.5, by feeding it examples of correct responses and explanations. After this tuning, it came close to ChatGPT 4.0 in accuracy and even surpassed it in providing answers’ that were similar to those of human professionals.

Both models still fell short, though, when it came to certain types of questions. While they did well reviewing securities transactions and monitoring financial market trends, the models gave more inaccurate answers for specialized situations such as determining clients insurance coverage and tax status.

Fairhurst and Greene, along with WSU doctoral student Adam Bozman, are now working on other ways to determine what ChatGPT can and cannot do with a project that asks it to evaluate potential merger deals. For this, they are taking advantage of the fact that ChatGPT is trained on data up until September 2021, and using deals made after that date where the result is known. Preliminary findings are showing that so far, the AI model isn’t very good at this task.

Overall, the researchers said that ChatGPT is still probably better used as a tool to assist rather than as a replacement for an established financial professional. On the other hand, AI may change the way some investment banks employ entry-level analysts.

“The practice of bringing a bunch of people on as junior analysts, letting them compete and keeping the winners – that becomes a lot more costly,” said Fairhurst. “So it may mean a downturn in those types of jobs, but it’s not because ChatGPT is better than the analysts, it's because we’ve been asking junior analysts to do tasks that are more menial.”

Journal

DOI

Article Title

Article Publication Date