AI-generated election predictions could influence undecided voters, affect voter turnout and create a misleading impression of certainty, according to the Fijian Elections Office, as a data analytics specialist demonstrates how AI and machine learning could be used to forecast Fiji’s next General Election.
Fijian Elections Office Manager Legal Mesake Dawai says artificial intelligence is already present in Fiji’s election environment, with political parties using AI-generated images and other tools to develop campaign messages.
This comes as Data Analytics Specialist David Elbourne has developed a 12-factor prediction model which uses political trends, previous election results, social media sentiment, leader credibility, party resources and voter patterns to assess possible election outcomes.
Speaking during a presentation at the University of Fiji Press Club, which is led by students studying Journalism and Media at the University, Elbourne says the model looks at factors including political leaders’ credibility, online visibility, controversies, funding and resources, the spread of candidates around the country, iTaukei support and previous election performance.
He stresses that the model is not an opinion poll and does not provide a 100 percent certain election result.
Insert: Elbourne on AI 100% 10TH AUG 26
Elbourne says it instead combines qualitative and quantitative data to identify trends and calculate probabilities based on available information.
He says there are several factors that cannot be accurately predicted, including final voter turnout, voters switching their support, the election date, political party alliances, future events, the impact of campaigns and how individuals eventually vote.
Elbourne says the model does not remove uncertainty but measures it, allowing analysts to assess what is known against factors that remain unknown.
However, Dawai says from the perspective of the Elections Office, the use of technology to predict election outcomes raise concerns about its potential impact on voters and the electoral process.
He says the Electoral Commission activated the campaign period on 25th May, and political parties are now allowed to campaign.
Dawai says while AI provides opportunities for innovation and is already being used by political parties, it needs to be used responsibly.
Insert: Dawai on Party using AI 10 Aug 26
He says the Elections Office would be concerned if AI or other predictive tools were used to publish claims about who would win the 2026 or 2027 national elections.
Dawai says Section 110 of the Electoral Act makes it unlawful for anyone to publish the results of election-related opinion polls during the seven days before polling day and on polling day until polling has closed at all polling stations.
He says this type of blackout period is not unique to Fiji, with similar restrictions used in other countries.
Dawai says election predictions and opinion polls could influence undecided voters who have registered but have not yet decided who they will vote for.
He says they could also affect voter turnout as some people may decide there is no point in voting if they believe a particular party is already going to win.
Dawai says another concern is that predictions could create a misleading impression of certainty, while statistical estimates can also be affected or manipulated.
He says the restriction is about protecting voters and the integrity of the final stage of the election.
Dawai stresses that the Elections Office is not against artificial intelligence, but its use during the electoral process must be responsible.
He says democracy is about people making their own choices regardless of how accurately others believe they can predict the result, adding the accuracy of any prediction can only be determined once the election outcome is known.
Meanwhile, Elbourne says the predictive framework could be updated regularly as new information becomes available and could be used by political parties, journalists, analysts, policymakers and academic institutions to monitor changing trends.
He says similar predictive models could also be applied outside elections in areas such as climate change, economic development, infrastructure projects and government policy.
Elbourne also highlighted the wider impact of artificial intelligence, saying around 60 percent of jobs are directly impacted by AI.
He says universities need to move beyond the traditional approach of students enrolling, attending lectures, graduating and entering the workforce.
Elbourne is calling for Fiji to develop what he describes as "high-value universities", where institutions use research, data, artificial intelligence and local talent to tackle complex problems facing the country and develop practical solutions.
He says universities should create their own intellectual property from these ideas and frameworks, while students and academic institutions should be willing to ask difficult questions about Fiji’s economy, governance, policies and politics.
Elbourne says Fiji has the talent and capability to participate in the AI revolution but needs the innovation, courage and drive to take advantage of the opportunities it presents.
He is encouraging universities, students, policymakers and other institutions to develop locally created frameworks and intellectual property that can help address Fiji’s challenges.
