Artificial Intelligence, or AI, is the new buzz term. It is the new buzz term not just for the elections, but for how it is projected to change the way the world of work will change, how jobs will be done, and the changing ways in which humans interact with technology. So, what is meant by AI ? – AI refers to highly developed computing systems that aggregate a large and diverse set of technologies that synthesise large amounts of information and data that mimic human capabilities of reasoning, problem solving and decision making through the processing, analysing and interpretation of data and information according to set of instructions developed through algorithms.
This is taking everyday tasks that people routinely do, as a matter of everyday fact and making computer applications, robots and mechatronics, capable of doing them instead. The development and application of AI tools and instruments is moving faster than has been projected and predicted. AI is considered the next set of technological advances , and just as with other technologies shaping society, it offers significant potential for good, but also and poses huge risks. As AI becomes more accessible, it can be used productively to:
- To help spot online harms.
- Help analyse large amounts of data and extract salient issues and common societal concerns that journalists can use to add to their reporting.
- Build and encourage more open, rights-based societies.
As South Africa heads into its 2024 elections, AI has the potential to take on new dimensions in shaping the political contestation and competition and the electoral process.
AI can be harnessed for:
- Voter Education: We can create AI chat bots that can help with voter education and explanations of the voting process.
- Voter Engagement: AI can personalise voter outreach, tailoring messages to individual preferences and concerns, thereby enhancing engagement and potentially, voter turnout.
- Detecting Misinformation and Dis-information With the proliferation of mis and disinformation, AI tools can help sift through vast amounts of data to flag and filter out potential misinformation, ensuring a more informed electorate.
- Security Measures: AI-driven cybersecurity can fortify the integrity of electronic voting systems against hacking and other digital threats.
- Data Analysis: AI excels in analysing voter data and could be used to model turnout and behaviour, offering valuable insights for campaign strategies for political parties and voter outreach for the IEC.
- Accessibility: AI can make voting more accessible through voice-activated systems and language translation services, breaking down barriers for those with disabilities or language differences.
At the same time AI can also be used by political and other actors with bad intentions. They can manufacture and proliferate mis and disinformation on an industrial scale, and manipulate images, voices and video in ways that make it almost impossible to distinguish from actuality.
Some potential ways AI could undermine the integrity of elections:
- Deepfakes and Synthetic Media: AI can create highly realistic fake videos or audio recordings, known as deepfakes, which can be used to spread false information or impersonate political figures.
- Social Media Manipulation: Algorithms can be used to generate or amplify disinformation on social media platforms, influencing public opinion and voter behaviour.
- Data Privacy Breaches: AI tools can exploit vulnerabilities in data security, leading to unauthorized access to sensitive voter information and election databases.
- Automated Bots: AI-powered bots can flood online discussions with spam or biased content, skewing public discourse and potentially swaying voter sentiment.
- Microtargeting: AI can analyse vast amounts of personal data to enable hyper-targeted political advertising, which may manipulate voters by exploiting their personal biases or fears. Further polarising communities.
- Voting Pattern Prediction: While its analysis could be used for the public good, AI can predict voting patterns and behaviours, which could be misused to disenfranchise certain voter groups or manipulate electoral outcomes.
- Cyberattacks: AI can enhance the capabilities of cyber attackers, making it easier to disrupt election infrastructure, including voting machines and voter management devices, including ballot counting systems. South Africa however, still relies on a physical ballot count which might mitigate the ill effects of cyberattacks on the count. The risk of cyberattacks corrupting other part of the electoral process is real.
- Psychological Profiling and the abuse of fake facts: AI can now be used to create detailed psychological profiles of voters, which could be exploited to craft persuasive and manipulative political campaigns. Persuasion (and a degree of manipulation) of voter sentiment is a part of the contest of political and policy ideas, and an integral part of robust campaigns. The danger posed by AI is that it may pass off false information, fake news or outright lies as reality and fact. Coupling fake news with the ability of AI models to psychologically profile and target voters, is enabled by technology and AI, something that not necessarily have been possible before its advent.
Practically, what does all of this mean for our election period in South Africa? As a member of the public it means that the risk of being misled by mis and disinformation is potentially greater. As the quality of AI improves it is becoming more difficult to distinguish real from synthetic or manufactured data and information. The ability to manufacture audio that sounds like a real person is already quite advanced. An example that was easier to spot, was a video of Donald Trump appearing to endorse a particular South African political party. The post was reported to Real411 , and the video can be seen here.
While the video is accompanied by some laughing emoji’s and the lip syncing isn’t very well matched to the words which seem to be articulated, it is evident that it is easy for a person to manipulate moving images, and life-like animations of leaders, public figures, celebrities and popular personalities to be purporting to be saying almost anything. This is direct manipulation of potential political endorsement and influence, and where it is used to make nefarious claims, it could have legal implications for the creators of such content.
Using AI generated content for the purposes of disinformation and manipulation is simply another sophisticated tool used to disseminate harmful content and manipulate and influence the public. In relation to the South African election, we have seen a few examples of AI generated content, like the posts highlighted above, but we haven’t seen an overwhelming amount of this content used for campaigning, although this is being investigated further. What we have seen, is manipulated content, and statements and claims that are simply not true, or are borderline incitement .
As the quality of AI improves it is becoming more difficult to distinguish real from synthetic or manufactured data and information.
An example of incitement related to election can be seen in this complaint reported to Real411.
The complaint is about a tweet directed at Herman Mashaba contains a threat of militant action against him and his political party, framed as a response to perceived attacks on the EFF and its leadership.
Although there is an abundance of potential positive applications of AI, one needs to be far more critical of content consumed online, and as a default, to ensure the verification of images and audio that appears on line.
AI can also be used for positive impact, enabling voters to have easier access to information and education material. One organisation that has used AI for this purpose, is Rivonia Circle. Rivonia Circle have developed Thoko the Bot.
Thoko is a bot that uses AI technology to answer questions and engage with users on information related to the national elections. After exchanging a few messages with Thoko, one is able to get information around the importance of voting in South Africa, where to vote, eligibility to vote, and how to register to vote.
Artificial Intelligence is a technology that is being adopted on a wider scale in South Africa, and although there is not one solution to address the use (and abuse) of it, it is clear that there is a need for more critical engagement with online content. In this context the media have an even more important and pronounced role to play in disseminating factual, timely and accurate information and interpretations of them – especially during an election period.
As a general rule, if online posts makes one feel anxious or angry, or fearful – this would often be an indicator of an attempt to mislead or misinform. Examine the source of the information, if it is about a senior leader saying or doing something and it isn’t a credible news source then check on a credible site before sharing it
Pause before you post – social media encourages users to share and move on but often its useful especially with controversial content to pause, have a cup of tea or something stronger – and then then check, verify if possible before posting. If you see something dodgy – report to Real411.org.za.
The South African Elections Weekly Briefs are produced through a partnership between The Electoral Institute for Sustainable Democracy in Africa, Media Monitoring Africa and the Institute for Justice and Reconciliation. The purpose of the partnership interventions is to strengthen peaceful and inclusive participatory electoral processes in South Africa.
