Laptop open with a display of AI search bar.

Certainly! #GenerativeAI and its Impact on Academic Writing (in Finance)

This paper investigates how the introduction of Large Language Models (LLMs) has affected academic writing in finance.

Authors: Thomas Walther, Marie Dutordoir

Read More: http://spkl.io/63320ABWym

𝗗𝗶𝗱 𝘆𝗼𝘂 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝘁𝗵𝗶𝗻𝗸 𝗖𝗵𝗮𝘁𝗚𝗣𝗧 𝗶𝗺𝗽𝗿𝗼𝘃𝗲𝘀 𝘆𝗼𝘂𝗿 𝘄𝗿𝗶𝘁𝗶𝗻𝗴?

We looked at the last 25 years of published articles in 34 finance journals. Can we see an effect of ChatGPT on the academic writing? We can! 𝗥𝗲𝗮𝗱𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗱𝗲𝗰𝗹𝗶𝗻𝗲𝘀 𝗱𝗿𝗮𝘀𝘁𝗶𝗰𝗮𝗹𝗹𝘆.

papers.ssrn.com/sol3/papers....

Thank you for sharing!

Woman typing on laptop surrounded by textbooks and notebooks.

This paper presents empirical evidence that, following ChatGPT's release in late 2022, the readability of #finance papers has declined significantly.
http://spkl.io/63323A88Zv
@twalther.bsky.social

👉 The working paper and its online appendix are available now on SSRN: papers.ssrn.com/sol3/papers....

We believe our findings are important for academics, editors, publishers, and research funding bodies, but also for the wider society in terms of accessibility to the ivory tower of science.

Lastly, we present evidence suggesting that a higher reliance on ChatGPT makes authors 𝗺𝗼𝗿𝗲 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗲 (more publications per year), but 𝗱𝗼𝗲𝘀 not 𝗵𝗲𝗹𝗽 𝘁𝗼 𝗶𝗺𝗽𝗿𝗼𝘃𝗲 publication 𝗾𝘂𝗮𝗹𝗶𝘁𝘆 𝗻𝗼𝗿 scholarly 𝗶𝗺𝗽𝗮𝗰𝘁.

Conversely, author teams from countries that place less emphasis on creativity, display lower skepticism toward technology, and have higher levels of moral ambiguity show a more pronounced effect.

𝗪𝗵𝗶𝗰𝗵 𝗳𝗮𝗰𝘁𝗼𝗿𝘀 𝗮𝗿𝗲 𝗶𝗻𝗳𝗹𝘂𝗲𝗻𝗰𝗶𝗻𝗴 𝘁𝗵𝗶𝘀 𝗱𝗲𝗰𝗿𝗲𝗮𝘀𝗲 𝗶𝗻 𝗿𝗲𝗮𝗱𝗮𝗯𝗶𝗹𝗶𝘁𝘆?

We find that author teams from primarily English-speaking countries do not show this drastic effect.

We then move on and document a decrease in the readability of finance articles since the advent of ChatGPT. Papers require 𝗼𝗻𝗲 𝗳𝘂𝗹𝗹 𝗮𝗱𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝘆𝗲𝗮𝗿 of education to read comfortably. The decrease of readability coincides with an increase of AI-associated words.

First, we run a set of simulations. We let LLMs generate introductions based on actual abstracts and results sections and then compare the generated introduction with the original one. On average, none of the LLMs is able to create a more readable version of the introduction.

𝗗𝗶𝗱 𝘆𝗼𝘂 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝘁𝗵𝗶𝗻𝗸 𝗖𝗵𝗮𝘁𝗚𝗣𝗧 𝗶𝗺𝗽𝗿𝗼𝘃𝗲𝘀 𝘆𝗼𝘂𝗿 𝘄𝗿𝗶𝘁𝗶𝗻𝗴?

We looked at the last 25 years of published articles in 34 finance journals. Can we see an effect of ChatGPT on the academic writing? We can! 𝗥𝗲𝗮𝗱𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗱𝗲𝗰𝗹𝗶𝗻𝗲𝘀 𝗱𝗿𝗮𝘀𝘁𝗶𝗰𝗮𝗹𝗹𝘆.

papers.ssrn.com/sol3/papers....

Among other topics, future research should address 𝙬𝙝𝙮 𝙁𝙍𝘼𝙨 𝙙𝙤 𝙣𝙤𝙩 𝙖𝙥𝙥𝙚𝙖𝙡 𝙩𝙤 𝙡𝙚𝙨𝙨 𝙛𝙞𝙣𝙖𝙣𝙘𝙞𝙖𝙡𝙡𝙮 𝙡𝙞𝙩𝙚𝙧𝙖𝙩𝙚 𝙥𝙚𝙤𝙥𝙡𝙚, who likely would benefit more than others from using FRAs and 𝙩𝙝𝙚 𝙞𝙣𝙩𝙚𝙜𝙧𝙖𝙩𝙞𝙤𝙣 𝙤𝙛 𝙇𝙖𝙧𝙜𝙚 𝙇𝙖𝙣𝙜𝙪𝙖𝙜𝙚 𝙈𝙤𝙙𝙚𝙡𝙨 𝙖𝙣𝙙 𝙂𝙚𝙣𝙚𝙧𝙖𝙩𝙞𝙫𝙚 𝘼𝙄 to enhance user interaction.

We review on the nascent research on FRAs. We identify two streams of literature:
(1) 𝗮𝘀𝘀𝗲𝘁 𝗺𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁, which focuses on designing FRAs and improving the functioning of these machine advisors, and
(2) 𝗯𝗲𝗵𝗮𝘃𝗶𝗼𝘂𝗿𝗮𝗹 𝗳𝗶𝗻𝗮𝗻𝗰𝗲, which investigates technology adoption and issues related to biased advice.

Financial Robo-Advisors (FRAs) enable households to participate in financial markets w/ a limited amount of money and w/o time or place constraints. FRAs can help investors overcome behavioural biases, but have disadvantages, like relying on a limited number of inputs and lacking individualization.

🤖 𝗙𝗶𝗻𝗮𝗻𝗰𝗶𝗮𝗹 𝗥𝗼𝗯𝗼-𝗔𝗱𝘃𝗶𝘀𝗼𝗿𝘀: 𝗪𝗵𝗮𝘁 𝘄𝗲 𝗸𝗻𝗼𝘄, 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝘄𝗲 𝗱𝗼𝗻’𝘁 𝗸𝗻𝗼𝘄!

I am happy to share my new working paper “𝙁𝙞𝙣𝙖𝙣𝙘𝙞𝙖𝙡 𝙍𝙤𝙗𝙤-𝘼𝙙𝙫𝙞𝙨𝙤𝙧𝙨: 𝘼 𝘾𝙤𝙢𝙥𝙧𝙚𝙝𝙚𝙣𝙨𝙞𝙫𝙚 𝙍𝙚𝙫𝙞𝙚𝙬 𝙖𝙣𝙙 𝙁𝙪𝙩𝙪𝙧𝙚 𝘿𝙞𝙧𝙚𝙘𝙩𝙞𝙤𝙣𝙨” co-authored with Mustafa Nourallah, Peter Öhman, and Duc Khuong Nguyen.

👉 papers.ssrn.com/sol3/papers....

Las Palmas de Gran Canaria

I am very happy to share that Tony Klein and I have co-edited a Special Issue on xAI in Finance, now published in 𝘍𝘪𝘯𝘢𝘯𝘤𝘦 𝘙𝘦𝘴𝘦𝘢𝘳𝘤𝘩 𝘓𝘦𝘵𝘵𝘦𝘳𝘴!

📄 Read the editorial: doi.org/10.1016/j.fr...
📚 Explore all articles: www.sciencedirect.com/special-issu...

𝗡𝗲𝘄 𝗽𝘂𝗯𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗮𝗹𝗲𝗿𝘁!

Forecasting realized volatility of crude oil futures prices based on machine learning

open access: onlinelibrary.wiley.com/doi/10.1002/...