How AI could transform Indonesia's election oversight

  Amid rapid technological advances, artificial intelligence (AI) is playing an increasingly prominent role across various sectors, including election oversight.

Once primarily used to improve workplace efficiency, the technology now offers much greater potential to help detect suspected violations, identify patterns of information manipulation, and provide early warnings of risks to democratic integrity.

This presents Indonesia with an opportunity to develop an election oversight system that is more responsive, accurate, and transparent.

In general elections involving millions of voters, thousands of polling station officers, and rapidly evolving digital information, manually verifying all incoming data is becoming increasingly challenging.

AI offers a new approach to election oversight, shifting the focus from relying solely on incoming reports and post-election reviews toward using data analytics to identify potential problems at an early stage.

Deputy for Technical Support at the Election Supervisory Agency (Bawaslu), Yusti Erlina, explained that AI could help transform oversight from a system focused on administrative reporting into one driven by data analysis and early warning mechanisms.

The technology can help identify unusual data patterns and flag suspected violations for further investigation.

However, one fundamental principle must not be overlooked: algorithmic anomalies do not constitute proof of a violation. Every suspected violation must be verified by human inspectors based on evidence, context, and applicable legal provisions.

This ensures that AI remains a decision-support tool rather than a substitute for human judgment and institutional authority.

Although algorithms can identify relationships and patterns across large volumes of data, determining whether a violation has occurred remains the responsibility of authorized institutions operating under established legal procedures.

Bawaslu has outlined three key functions of AI in election oversight: guardian, assistant, and reminder.

As a guardian, AI can help identify suspected document manipulation, synthetic content such as deepfakes, and coordinated information distribution patterns that warrant further investigation. This capability is increasingly important as images, audio, and videos become easier to manipulate, making it more difficult for the public to distinguish authentic information from altered content.

As an assistant, AI can categorize public reports, summarize information, and help prioritize investigations based on available data. This enables election supervisors to allocate their time and resources more effectively.

Meanwhile, as a reminder, AI must generate analyses that are traceable, verifiable, and accountable. Election officials remain responsible for examining evidence, understanding the surrounding context, and ensuring that decisions comply with applicable laws and procedures.

Together, these three functions illustrate a significant shift in the election oversight paradigm. AI is no longer viewed merely as a tool for speeding up administrative work but also as a means of improving accuracy, consistency, and the ability to anticipate potential problems.

Preliminary indicators

AI systems can be designed to identify unusual similarities between documents, detect duplicate reports, and recognize digital content being distributed through potentially coordinated networks.

Rather than automatically establishing that a violation has occurred, these findings serve as preliminary indicators requiring further examination.

For these benefits to materialize, data quality remains essential. AI systems operating on incomplete, inaccurate, or biased information risk producing misleading conclusions. Therefore, transforming election oversight through AI must begin with improvements in data governance.

Currently, Bawaslu relies on several digital tools. Its prevention reporting form is accessible through a web-based system designed to support internal operations.

For instance, the Siwaslu system is used during general elections, while Siwaslih supports oversight during regional elections. The existence of separate applications highlights the need for greater integration to prevent election oversight data from becoming fragmented across independent systems.

The Integrated Oversight Single Data initiative provides a foundation for more systematic AI deployment. Integrating these systems would allow information to be consolidated, cross-checked for consistency, and analyzed using unified standards.

The proposed Participatory Oversight SuperApp also offers the potential to establish a unified platform for handling violation reports, voter education, and final voter list verification. Within an integrated framework, public reports could be categorized according to location, time, and type of issue.

AI could then help identify emerging patterns requiring the attention of election supervisors, while preserving the public's right to a fair review process.

Blockchain technology is also being explored as part of these efforts. Under a hybrid approach, data processing and storage could take place off-chain to maintain operational efficiency, including for vote tallying at polling stations.

Meanwhile, on-chain mechanisms could be tested to record data audit trails, making changes to information easier to trace and inspect. Although blockchain cannot guarantee the accuracy of data at the point of entry, it can significantly improve traceability when properly designed and implemented.