CSIR builds tools to spot fake election videos and audio, and wants partners to help

The Council for Scientific and Industrial Research (CSIR) has called on partners to work with it on using technology to combat misinformation, saying artificial intelligence is changing how information is created, shared and consumed.

Sophisticated AI-generated videos, audio and images are making it harder to tell authentic content from misleading or fabricated material, the CSIR said, raising concerns about misinformation and disinformation, particularly during elections.

Speaking at a virtual media roundtable on Wednesday, the CSIR’s acting research group leader for natural language processing and principal AI specialist, Avi Moodley, said organisations fighting misinformation and fake news cannot afford to work in isolation.

“The challenge is enormous. Whether it is data, infrastructure, resources or funding, all these elements can make a significant difference to our coordinated efforts. We would welcome your participation in strengthening and expanding this network,” Moodley said.

The roundtable, hosted by the CSIR with the Government Communication and Information System and other stakeholders, examined the challenges and their implications for the electoral information environment.

Moodley said the volume of content makes it difficult for people to judge what they receive. “You do not know what is real and what is AI-generated. If we, as researchers, experience this fatigue when analysing information and data, can you imagine what ordinary South Africans are experiencing when they are overwhelmed with information and are uncertain about what is real or fake?”

She said advances in AI have, to some extent, industrialised the spread of misinformation, disinformation and malinformation, together known as information disorder. “AI can be used to create synthetic content, develop and amplify emotionally charged narratives, and produce material that could potentially manipulate perceptions in specific contexts.”

Moodley said the flooding of the information ecosystem has created an environment unlike previous elections. “We have not seen this phenomenon in previous elections to the extent that we are likely to experience it in the future. Our response, as organisations working to combat misinformation, therefore needs to be coordinated because we cannot address these challenges in isolation.”

According to the CSIR, it follows a set process during each election. It identifies keywords that shape election-related discourse, collects information from social media platforms, and enriches it using its natural language processing services and additional data. That information is processed through its misinformation and unrest detection systems to identify potential warning signals, and can be used for network analysis and shared with partners who need access to the data.

The findings are incorporated into dashboards the CSIR will showcase throughout the election period, and the council aims to share the information widely with partners and through public dissemination platforms, including the Southern African Development Community.

The CSIR has developed the Informed Fake News Advisor Project to evaluate information, compare it with credible sources and trace its origin, with the aim of providing indicators that help identify and filter potential misinformation. Moodley said the technology is not necessarily intended for everyday use by ordinary South Africans but could benefit fact-checkers working through large volumes of information to find content that needs further investigation and reporting.

The council is also developing an information triage system that takes available information, together with variables and inputs generated by the Informed Fake News Advisor, and helps prioritise content for further assessment. Moodley said this would help the CSIR prioritise information needing urgent attention in time-sensitive situations such as elections.

“This technology is still a work in progress. We are currently looking at using the upcoming elections as an opportunity to test some of these technologies so that we can strengthen our capabilities as we move from the local government elections towards the next national elections,” she said.

The CSIR is also developing technology to search audio content, extending its work beyond text. Moodley said efforts to combat misinformation in the country are largely limited to processing text, and that audio presents challenges arising from South Africa’s linguistic diversity, including the wide range of languages spoken and the limitations of converting audio into text for analysis.

“We are therefore developing technology that bypasses the need for this conversion and allows us to search audio content quickly and efficiently. We anticipate that this could be a game-changer in addressing multimodal content, which is becoming increasingly prevalent with the rise of AI,” she said.

This report is based on the official account published by SAnews (South African Government News Agency): https://www.sanews.gov.za/south-africa/csir-seeks-partners-combat-ai-driven-misinformation