Herkulaas MvE Combrink, University of the Free State
Social media has reached more than half (63.9%) of the world’s population since it got started in 1996. Social network platforms grew from 970 million users in 2010 to 5.41 billion in July 2025. The average social media user engages with between six and seven platforms. The average person spends two hours 21 minutes on social media per day.
All this time on social media shapes people’s perceptions, influences emotions and fuels anxieties, but its impact on social stress remains difficult to measure. Herkulaas MvE Combrink is a specialist in computational infodemiology, a field which studies the spread of information in digital spaces, at high volumes and with high degrees of uncertainty. The discipline draws on artificial intelligence, public health, and natural language processing. He unpacks his research and explains the Social Stress Indicator, a computational tool he devised for quantifying social stress in real time. It can provide an early warning of how people are feeling about what they are seeing online.
|
|
What is social stress?
Social stress is the tension people feel when their social environment becomes uncertain, demanding, or conflicted. It goes beyond personal anxiety, reflecting how communities collectively react to issues like inequality, misinformation, or rapid change.
In a societal context, social stress describes the shared emotional pressure that spreads through networks, shaping public mood, cohesion and behaviour.
|
|
|
|---|
In the context of an infodemic (a flood of information during a disease outbreak), misinformation and uncertainty becomes the primary indicator driving social stress.
What is the Social Stress Indicator and why did you develop it?
It is one of the few algorithmic stress trackers tailored for the digital age. The data it uses is primarily text and search volume indices.
What makes this indicator unique is that it can be adjusted for social media platforms, enabling the real-time tracking of social stress per topic of concern. A topic of concern may be a specific conversation, like vaccines or public elections.
The indicator can therefore be used for analytical insight, as a means to analyse retrospective data, or to assist in informing policy by quantifying social stress over time.
Social media is no longer just communication – it’s a battlefield of emotions, fears and societal pressure, as seen by the impact of misfluencers (opinion leaders who disseminate misinformation about health) and the social media perspectives on COVID-19 vaccines. These misfluencers had a high likelihood of both confusing and misleading their followers on social media.
Social media can act as an “echo chamber” for belief systems to thrive in. An echo chamber reinforces stereotypes, making it the perfect incubator for misinformation spread.
My research has shown that these echo chambers can be studied, if the correct indicators are used. Here I mean indicators like search volume indices, sentiment analysis, polarity scores, subjectivity scores, social volatility score, sceptic score, and ultimately, the social stress indicator.
I was part of a study that assessed 150,423 unique social media posts across various groups on the social media platform X, related to COVID-19 in South Africa. We found that the majority of posts were related to access to health, mismanagement of COVID-19 and vaccine hesitancy. We compared the findings to the vaccine roll-out and noted a trend between a decrease in public sentiment and vaccine uptake. In other words, when people started saying negative things about vaccines, vaccination uptake and programmes slowed down.
These figures and insights required extensive research to quantify and investigate because of the contextualisation and volume of data to analyse to draw these conclusions. Although we could clearly see a decrease in vaccine uptake, we investigated the discourse in relation to the decrease.
I built the indicator to quantify this chaos. Instead of working with fragmented analysis and struggling with the combination of various data sources and independent analysis, the social stress indicator simplifies this process into a number that’s easy to track.
Traditional mental health surveys are too slow, and online toxicity is too fast because misinformation spreads too rapidly to track. This means that when there is an infodemic, its impact on what people are thinking and feeling becomes too challenging to measure in real-time. The indicator steps into this gap, offering a quantifiable pulse of online distress.
How does it work?
It works like a digital thermometer for a specific topic online, using what people say on social media, and then fuses three ingredients:
- sentiment analysis, to detect emotional negativity
- subjectivity, to measure personal intensity in expression
- information-seeking behaviour, to track public concern.
These go into calculating a final Social Stress Indicator score. Think of it as a stress composite where negative sentiment raises the alert, subjective opinion adds weight, and spikes in search traffic reflect societal urgency.
A low Social Stress Indicator means calm waters. A high Social Stress Indicator? Digital panic may be brewing.
We analysed three social media datasets that people rated for how stressful they felt, based on negativity, anxiety, engagement, misinformation, and help-seeking behaviour. We then applied the Social Stress Indicator to the same data to compare human perception with data-driven patterns. The results showed that while the overall trends matched, the intensity of stress varied across different contexts.
Why will it be useful?
Imagine a world where governments and health systems can spot societal breakdowns before they go viral. The indicator offers exactly that. As an early-warning system, it can detect digital tremors before they become “social earthquakes”, whether it’s vaccine hesitancy, misinformation spikes, or panic over policy changes.
It provides the missing metric, the emotional cost of communication. If social stress is not tracked, misfluencers gain traction and misinformation can spread. Another concern if social stress is not managed is the potential for social unrest and public crisis. It’s therefore invaluable for digital governance, crisis response, mental health monitoring, and platform design to track and respond to social stress. The indicator doesn’t just observe, it alerts.
What are its flaws?
It needs to be more sensitive. When social stress soars, it tends to stay calm. That’s because it was designed for general trends and not emotional eruptions. For more sensitivity, more research is required.
It also assumes equal weight for sentiment, subjectivity and curiosity, which dilutes urgent signals. For instance, during a vaccine misinformation surge, posts expressing fear or anger (sentiment) may rise sharply, while curiosity-related searches only increase slightly. Giving both equal weight could underestimate the level of public distress.
Plus, it struggles with sarcasm, trolling, or coded hate, which are things that humans detect instantly.
The model also depends on synthetic validation (not real-world chaos), which means its real-time usefulness might falter when tested against messy reality. More research is required in this domain to make the indicator more robust. Until it adapts dynamically, its risks are more like a weather report than a storm warning for social stress on digital media. As the Knowledge Mapping Labs at the University of the Free State, South Africa, grows, this work will continue.
Herkulaas MvE Combrink, Senior lecturer/ Co-Director, University of the Free State
This article is republished from The Conversation under a Creative Commons license. Read the original article.
