AI Report on Southern Africa http://misa.org AI for personalised advertising and microtargeting Personalised advertising is a legal way to transmit political campaign and voter information, but it may also be used for manipulation through micro-targeting. The development of user profiles is required for individually and directly addressing voters. This construction of personality profiles is mainly employed for personalised advertising and forms part of the dominant digital platforms’ core business strategy. These advertising skills contribute to the possible risk of manipulation in politics and elections using the same tactics as Cambridge Analytica. Targeted and personlised advertising has been actively used in Zimbabwe elections by the two main political rivals Zanu PF and Citizens Coalition for Change (CCC). In general, a low efficacy rate is enough to make a difference especially in countries like Zimbabwe where the winning margin between rival political parties is small. Thus, even if only a few out of thousands of individuals respond to the advertisement, it is worthwhile. It is worthwhile because even a small shift in voter preferences can significantly impact election outcomes. The influence of AI systems and electoral content control is frequently only evaluated in retrospect. Our initial observations of the Zimbabwean elections in 2023 indicate that there was a substantial quantity of basic disinformation and fake news peddled through micro target adverts on social media. As we have seen in Zimbabwe, there are numerous potential negative implications of AIdriven applications on voter information and public opinion formation. Some of them are most likely already influencing public opinion. AI to Counter Biased Content Although AI negatively affects public discourse, it can also enhance media content and factchecking. Artificial intelligence can assist in detecting biased material and provide alternative coverage. Typically, but particularly in elections, there is a proclivity for biased reporting. The phenomenon of media bias stems from the deliberate selection of specific vocabulary and subjects, known as “framing,” which presents information from a particular perspective. Due to the vast volume of news in digital media, it is no longer feasible for humans to manually observe, and fact-check online information. Due to this rationale, platform operators depend on the utilisation of automated systems that provide swift identification and evaluation of media bias. If required, these analyses can also be used as a foundation for counter-information. Browser plug-ins can provide users with tools, such as supplemental links containing balanced information, to enhance their ability to assess biased material while forming political opinions. AI-powered web services can offer counter-arguments to address skewed representations caused by algorithmic filter bubbles. Social bots can also aid in combating inaccurate and biased journalism. They can automatically distribute authenticated information and, if needed, engage in interactive dialogue to address inquiries within the realm of elections and the broader formation of political opinions. 33