Indirect aggression in Indian conversations on social media platform X is most common during major social, financial, sporting and political events, business school researchers have found after they used machine learning models to examine patterns of aggressive communication, analysing more than 1.31 lakh posts.
The study, titled “Indian Aggression Detection through Multiple ML Models from Twitter Data”, was conducted by researchers at Goa Institute of Management (GIM) and Goa Business School. The findings have been published in renowned journal “Advances in Consumer Research”.
The study explored whether machine learning could help to identify different forms of aggression in Indian X posts and how aggressive behaviour varied across emotionally and politically charged events.
The researchers classified the posts into three categories — Overtly Aggressive (OAG), Covertly Aggressive (CAG) and Non-Aggressive (NAG).
“The findings indicate that OAG responses among Indians remained relatively low across the incidents in four domains studied, ranging from 11 per cent to 37 per cent.
“Aggression levels also showed a general declining trend over time following each incident, although the rate of decline varied considerably across the different events,” Lakshmi Vishnu Murthy Tunuguntla, Professor at Goa Institute of Management, told PTI.
The study found that indirect aggression was common across several events.
“Open aggression was the highest during India’s T20 World Cup loss, that was at 37 per cent. Aggression generally declined over time, though the pattern varied across different events. Online reactions differed depending on the event and its duration,” he said.
According to P Balasubramanyam, Associate Professor in GIM’s Big Data Analytics Department, the research highlights the distinction between sentiment and aggression.
“While sentiment analysis broadly categorises communication as positive, negative or neutral, aggression detection focuses on harmful or hostile language, which may not always correspond with negative sentiment,” he said.
“The study contributes to the application of natural language processing and machine learning in computational social science, with potential implications for online moderation, responsible communication policies and further research into digital behaviour,” he added.
The study demonstrates how machine learning can be leveraged to detect and interpret public aggression in complex, multilingual environments like India’s digital landscape. Beyond classification, the research provides insight into how aggression manifests across different societal issues and over time.
“The findings have practical implications for improving online moderation systems, guiding responsible communication policies by the government, and informing future research into the psychology and sociology of digital interactions.
“By focusing on local context and diverse event categories, the research has made a contribution to computational social science and paves the way for more culturally attuned AI applications,” Balasubramanyam said.
