The South African Special Risks Insurance Association (SASRIA) is using artificial intelligence (AI) to analyse social and behavioural signals in an effort to forecast potentially destructive events.
This was revealed by Muzi Dladla, executive manager for stakeholder management at SASRIA, speaking during a panel discussion at the Social Media Summit.
The event was convened by Decode, the Pan-African strategic communications firm, in partnership with Johannesburg Business School and endorsed by the Public Relations Institute of Southern Africa.
Dladla explained the state-owned insurer, in collaboration with the Council for Scientific and Industrial Research (CSIR), is combining AI with human intelligence and social-science expertise to better understand the conditions that can precede protests, unrest, national disasters and damage to property and infrastructure.
SASRIA has a social mandate to cover risks excluded from traditional insurance, including losses arising from events such as civil unrest and national disasters.
Rather than waiting for an event to occur and responding afterwards, the objective is to identify signals early enough to improve preparedness and decision-making.
A major looming protest, for instance, is an event that can be pre-empted through social media analysis, he pointed out.
Dladla said this makes understanding social behaviour particularly important because protests are constitutionally protected, while the factors driving them can be complex and may have destructive consequences.
“The approach combines social-science analysis with AI, drawing information from social media and other data sources. These signals are then triangulated to produce correlated predictions that can help SASRIA understand what may be developing, weeks before the event happens,” he explained.
Hybrid approach
“We cannot necessarily predict exactly when people are going to destroy property, but we use AI to draw as much info as we can about what will likely happen. We are planning to create more models so that we look at it from a social science perspective.
“The objective is not to replace human judgement with algorithms. Instead, SASRIA sees AI as a form of augmented intelligence, where machine-based analysis is combined with human intelligence, contextual knowledge and subject-matter expertise.”
This hybrid approach is particularly important because AI can identify patterns across large volumes of information, while humans remain responsible for interpreting those signals and determining what they mean in a particular social context, he noted.
“AI cannot independently apply ethics in the same way humans can, while context and the dynamic nature of human behaviour remain difficult for machines to fully understand. AI systems also learn from data generated by people, meaning existing biases can be carried into their outputs.”
Dr Christopher Mahlathi, a CSIR engineer who participated in the panel discussion, said the partnership between the CSIR and SASRIA involves building models around complex systems.
The CSIR is working with SASRIA to build models that can make sense of complex patterns by bringing together signals from multiple data sources.
“A key part of the CSIR's role is combining established scientific modelling techniques with the ability of AI to process information more quickly.
“The technology is not designed to tell SASRIA what action to take once a potential risk is identified. Instead, AI helps establish relationships between events and the signals surrounding them, with the resulting models providing information that humans must interpret and act upon,” he explained.
Mahlathi said the underlying techniques have been used for years, but AI allows researchers to build and run predictive systems faster and analyse larger volumes of information.
Share
