Students outperform ChatGPT on accounting exam, score 76% to Chatbot's 47%

Students performed better on accounting exams than ChatGPT, a chatbot developed by OpenAI, according to research.Despite this, they praised ChatGPT, calling its performance "impressive" and a "game changer that will change the way everyone teaches and learns - for the better," reported PTI.

In a bid to determine how well OpenAI's technology will do on accounting examinations, researchers from Brigham Young University (BYU), US, and 186 other universities conducted this study. Their research was published in the journal Issues in Accounting Education.

The average student score on the accounting exam used by the researchers was 76.7 per cent, while ChatGPT received a score of 47.3 per cent.

While ChatGPT was shown to outperform students on 11.3 per centof the questions, doing particularly well on accounting information systems (AIS) and auditing, it was discovered that the AI bot performed worse on tax, financial, and managerial tests. The reason for this, according to researchers, may be that ChatGPT had difficulties with the mathematical operations necessary.

Further research revealed that the AI bot, which use machine learning to produce natural language writing, performed better on true/false and multiple-choice questions (69.5per centright), but less well on short-answer questions (between 28.7per centand 39.1per cent).

In general, the researchers found that ChatGPT had difficulty responding to higher-order questions. In fact, it was discovered that ChatGPT occasionally offered binding written justifications for wrong responses or offered multiple approaches to the same subject matter.

Importantly, researchers discovered that ChatGPT occasionally made-up facts. For instance, when providing a reference, it generated a reference that appeared authentic but was entirely false. Sometimes neither the work nor its authors were even real.

ALSO WATCH |Elon Musk again sounds warning bells on dangers of Artificial Intelligence to humanity | Details

Additionally, it was observed that the bot would divide numbers improperly or would add two numbers to a subtraction problem.

David Wood, the study's principal author and a professor of accounting at BYU, wanted to add to the heated discussion about how models like ChatGPT should be included into education. He therefore sought out as many professors as possible to participate in the study.

His social media co-author search went viral: 25,181 exam questions from classroom accounting were contributed by 327 co-authors from 186 educational institutions across fourteen different countries.

In order to provide ChatGPT with an additional 2,268 textbook exam bank questions, they also engaged BYU undergraduate students. The true/false, multiple choice, and short answer questions ranged in difficulty according to type, covering topics such as AIS, auditing, financial accounting, managerial accounting, and tax.

(With inputs from agencies)

WATCH WION LIVE HERE: