close
Thursday October 03, 2024

Will AI play a role in 2024 polls?

"We expect generative artificial intelligence to have a bigger impact politically than on business in short term," report says

By Aimen Siddiqui
September 29, 2023
A visitor watches an AI sign on an animated screen at the Mobile World Congress, the telecom industry’s biggest annual gathering, in Barcelona. — AFP/File
A visitor watches an AI sign on an animated screen at the Mobile World Congress, the telecom industry’s biggest annual gathering, in Barcelona. — AFP/File

KARACHI: The potential misuse of AI can be a cause for concern for countries preparing for elections in 2024, says a report published by the Economic Intelligence Unit (EIU).

The report, titled ‘Why AI Matters’, adds: “we expect generative AI to have a bigger impact politically than on business in the short term. With elections in the US, the UK and India in 2024, the accessibility of ChatGPT and its competitors will drive the creation of fake content, be it text, imagery, audio or video.”

Pakistan is likely to have its elections in 2024. Haseem uz Zaman, a journalist and fact-checker associated with Soch Fact Check in Karachi shares: “AI is a nascent field in Pakistan. People are still exploring the AI tools; the same is the case with the social media teams of different political parties. It is no secret that the former ruling party is one of the first and fastest players in the social media game and, therefore, it comes as no surprise that they were the first ones to hop on the AI ‘bus’. However, in my opinion, it is not currently a major threat at least ahead of the upcoming elections.”

Digital rights activist and founder of Media Matters for Democracy (MMFD) Asad Baig, however, shares: “Generative AI apps and algorithms are becoming much more accessible and popular, and thus, the likelihood of politically motivated generative imagery and content is potentially high. Considering the overall state of media and digital literacy, gender and religious prejudices, and political polarization, the impact of such disinformation could potentially be catastrophic.”

“But an important point to note here,” he adds, “is that the use of AI is not limited to generative imagery alone. The algorithms of social media platforms are largely dependent on machine learning and AI to promote and push their content, and Facebook Papers and other revelations have substantially proven that the algorithms push ‘hateful content’ the most for traction. Thus, the algorithms can potentially add fuel to fire in a charged political environment.”

Dr Muhammad Shareh Qazi, assistant professor at the Department of Political Science, University of Punjab, believes that gauging whether AI will have an impact on elections is a little complicated. He says that political parties “can always convince people by using various pieces of information that suit their needs. I think propaganda is always done by political parties all around the globe to get people to vote for them.” But an important point is: “will there be some AI-generated disruptions that would hamper the electoral process once people get into the balloting system?” Regarding this, he thinks “we do not have the technology at our disposal right now that could make this happen. However, if someone claims that the system was hacked in favour of a particular candidate, we have a proper system to deal with such allegations.”

Pakistan’s First AI-generated Image Case

In May 2023, the then information minister Marriyum Aurangzeb accused PTI supporters of using AI-generated images to mislead people, Haseem says that “ it was the first AI-generated content from Pakistan to be fact-checked and Soch Fact Check was the first outlet to debunk the claim right away. As for the government response, fact-checking on a governmental level has been and always will remain partial. While it was fascinating – and funny – to see the then-minister waving the poster in a press conference, I don’t think it was anything more than political point-scoring. It is the job of journalists and fact-checkers to do the reporting, fact-checking, investigating, and debunking; if done by the government, it loses its meaning.”

Asad adds: “It’s important that the government and policymakers understand the correct terminologies, and as opposed to a knee-jerk reaction towards the potential digital harms, consider a more cohesive, multi-stakeholder approach that isn’t grounded in political censorship or petty political gains.”

“Having said that, it has been demonstrated many times over that organized disinformation and hate speech is very real, and could create a lot of harm, and various political parties, including the PTI, have used it for political gains.”

Prof Shareh thinks that it is essential to see such incidents on the basis of two principles. “If the government feels that using AI-generated images to create misinformation or disrupt a certain process is a risk to national security, or security in general, the government is, in my understanding, mandated to carry out actions against such AI imagery.” But, he explains, if the information minister says that it was done to mislead the public or create a false narrative, it is important to see this as a remark made by a member of a political party. “As far as election campaigns are concerned, anything that acts as a force multiplier is good. ‘False imagery’ itself may also be good for only campaign purposes, but if it is something based on misrepresentation which is a crime or an offence, then the source of that image must be held accountable.”

He adds that if parties use AI-generated images or text to attract the crowd, this may be called ‘unethical’, but for somebody to do it, they are within their rights to do so. “The other party – as a contender – has the right to nullify whatever is being said. Such use is less of an argument on laws and more on how political parties react.”

Fact-checking AI in Pakistan

Haseem acknowledges that flagging AI content has its challenges: “AI-generated content is often put through other existing non-AI tools such as Adobe Photoshop to refine it and that hampers with identifying whether it is AI-generated or not. This becomes a hassle in Pakistan because South Asian facial features are different from what AI tools appear to have been trained on and, when trying to reverse-search or regenerate similar content, the results are not the same, which means that the process to fact-check them becomes even more challenging. There is also the issue of online detection tools that have been significantly inconsistent in flagging AI content.”

He says that the other issue is that “people whose AI content goes viral stay very lowkey and then act defensive when asked just for how they came up with the said content.”

Social Media as a Tool for Political Parties

On the presence of political parties on social media, Asad shares: “I feel each political party, even the religiopolitical groups, has a strong social media presence in its own way. There is, obviously, a question of effectiveness, and some might be better than others, but all political groups are trying their best to use social media platforms for political campaigning, and in most cases, it is working for them.”

Prof Shareh adds that for the urban population, social media plays a big role. Even in rural areas, social media has some impact. “Given that a very sizable population of Pakistan has access to the internet, if the 2024 elections are going to be “elections of perception”, I do agree that political parties will end up investing in social media platforms.”

The EIU report mentions that some governments are looking at regulating AI. Regarding this, Prof Shareh believes that the question of such regulation is irrelevant “unless we enter a phase where elections are held electronically.” He explains that if someone says that the country’s RTS system was hacked into by an artificially intelligent sentient system or some hacker was able to use AI to hack the system and that the same thing will be repeated in 2024, “then, yes, systems must be insulated against AI as well, and there has to be some security measures to keep these things in check.”

“But until that happens, regulating AI in a situation where elections are conducted through ballots would not be important.” He also adds that as far as AI’s impact on election campaigns is concerned, “you cannot restrict modes of dissemination of information; you can surely manage them, which the government is already doing by introducing various laws against electronic dissemination of information. Beyond that, it should not do much because then it would be authoritarian of them to dictate what is okay and what is not okay.”

Asad adds: “As for regulations on AI, it’s very complicated, as opposed to creating a national regulatory framework, which is very likely to be used for political censorship. Since the harms that AI creates aren’t limited to, or generated from, Pakistan, a global approach is needed to counter them. There are various global efforts that are currently going on to address it, including the ones led by the UN.”