AI and the Future of Election Oversight
Jakarta (ANTARA) - Artificial intelligence (AI) is beginning to play a role in election oversight. Technology that has so far been widely used to improve work efficiency now has greater potential: helping to detect suspected violations, identify patterns of information manipulation and provide early warnings about risks that could affect the integrity of the democratic process.
This change offers Indonesia an opportunity to build an election oversight system that is more responsive, accurate and transparent.
In a democracy involving millions of voters, thousands of officials and fast-moving flows of digital information, people’s ability to check all data manually is increasingly limited.
AI offers a new approach. Oversight need no longer rely solely on reports received or checks conducted after an event has occurred; data analysis can instead be used to identify indications of problems at an early stage.
Yusti Erlina, SH, Deputy for Technical Support at the Indonesian Election Supervisory Agency (Bawaslu), explained that AI could shift oversight from administrative record-keeping towards an analysis- and early-warning-based system.
The technology can help identify unusual patterns in data and flag indications of suspected violations for further examination.
However, a fundamental principle must not be overlooked: an anomaly detected by an algorithm is not proof of a violation. Supervisors must verify every indication by considering the facts, context and applicable law.
This principle positions AI as a decision-support tool, not a replacement for human authority.
Algorithms can identify relationships between data points, but determining whether a violation has occurred remains the responsibility of the authorised institution, acting in accordance with legal procedures.
Bawaslu has outlined three roles for AI in election oversight: guardian, assistant and reminder.
As a guardian, AI can help identify suspected document manipulation, synthetic content such as deepfakes, and patterns in the spread of information that warrant examination.
This capability is relevant as images, audio and video become increasingly easy to manipulate, potentially making it difficult for the public to distinguish authentic information from altered content.
As an assistant, AI can group public reports, summarise information and help prioritise checks based on available data. This could enable supervisors to allocate their time and resources more effectively.
As a reminder, AI must produce analyses that can be traced, tested and held to account. Supervisors remain responsible for examining evidence, understanding the context and ensuring decisions comply with their authority and legal procedures.
These three functions illustrate an important shift in the approach to oversight. AI is not simply a means of speeding up work, but can also help improve the accuracy, consistency and foresight of supervisory institutions.
Early indications
For example, an AI system could be designed to find unusual similarities across documents, identify duplicate reports or detect coordinated patterns in the spread of digital content.
Such findings could provide an initial lead for further investigation, rather than an automatic conclusion that a violation has occurred.
For these benefits to materialise, data quality is fundamental. AI working with incomplete or inaccurate information, or data containing bias, risks producing incorrect conclusions. The transformation of oversight through AI must therefore begin with improvements to data governance.
Bawaslu currently has several digital tools. Prevention forms are used through an internal web-based system.
Siwaslu is active during elections, while Siwaslih is used during regional elections. Jarimu Awasi Pemilu, meanwhile, cannot yet be relied on as a service channel because it is inaccessible.
The existence of these various applications points to the need for integration, so that oversight information is not scattered across systems operating independently.
The proposed Integrated Oversight One Data system offers a foundation for more systematic use of AI. Integration would allow information to be consolidated, checked for consistency and analysed according to common standards.
The Participatory Oversight SuperApp concept also opens the possibility of providing a single service for reporting suspected violations, election education and checking the permanent voter register.