Fourteen objections, one question: Who can challenge electoral roll’s algorithm?

For nearly two decades, India did not undertake a nationwide intensive revision of its electoral rolls. Instead, the Election Commission largely relied on annual and pre-election special summary revisions, making additions and deletions to existing rolls rather than preparing electoral rolls afresh.

The latest investigation by The Indian Express has now added another layer to the debate: questions raised from within the Election Commission itself.

In late September, The Indian Express reported that Election Commissioners Sukhbir Singh Sandhu and Vivek Joshi had formally objected at least 14 times over ten months to decisions concerning voter registration, deletion, restoration, appeals and access to electoral-roll data.

In one case involving changes to Form 6, used by new voters, Sandhu described the changes as “unauthorised and illegal”. The report also detailed concerns over the “gradual centralisation” of the electoral-roll database.

The Election Commission responded that differing views are a normal part of deliberation in a multi-member constitutional body and maintained that all final decisions, including SIR, were unanimous.

That distinction matters. The existence of objections does not by itself establish that SIR is flawed.

But the 14 objections are significant because they were not confined to one administrative detail. They touched registration, deletion, restoration, appeals, changes to voter forms and access to the systems that hold electoral-roll data. The disagreement therefore raises a broader question about the architecture of the exercise.

When a central system flags a voter as absent, unmapped or discrepant, who gets to decide whether the machine or the voter is wrong?

For the Northeast, this is not an abstract question.

The machine is useful. The mismatch is the problem

The Election Commission has a legitimate administrative reason for using a common digital architecture.

ERONet, now operating within the wider ECINet system, processes electoral forms and helps officials manage voter records. The system can standardise applications, track cases and identify similar or potentially duplicate entries. The ECI has also described its digital systems as tools for managing electoral rolls across constituencies.

SIR is not simply an algorithm for deleting names. The process includes enumeration by Booth Level Officers, draft rolls, claims and objections, notices and hearings.

The difficulty begins when a computerised system turns an old record into a present-day question.

The Commission’s voter search system itself advises users to try spelling variations because names may differ between old and current electoral rolls. That small instruction captures a larger problem in a country where names move between languages, scripts and transliterations.

A mismatch is not necessarily fraud. It is evidence that something needs checking.

And checking is where the human machinery of an electoral roll matters.

What “44 per cent” looks like on the ground

The abstraction becomes concrete in Delhi’s Badarpur constituency.

In September, The Indian Express reported that around 87,000 electors, or 44 per cent of Badarpur’s electorate, were being served notices because their details could not be linked to the 2002 electoral roll or because anomalies had been found in their enumeration forms.

At Panchsheel Public School, Booth Level Officer Sumit Bhati had already conducted 36 hearings by around 2.30 pm on one day. He had been allocated 1,000 electors, around 240 of whom were expected to receive notices, mostly over discrepancies such as spelling mismatches.

The ERO office told the newspaper that voters could submit documents to BLOs at home or at polling stations, while final approval would remain with the Electoral Registration Officer.

Three kilometres away, 23-year-old Anita arrived with her own education certificates and those of her father. She had received a notice because she could not establish a link to the 2002 electoral roll.

“These people should have told us at once what documents are needed. Every time they ask for different documents,” she told The Indian Express.

Her experience illustrates the documentary burden SIR can place on voters whose present records do not map neatly onto a two-decade-old electoral roll.

The system has a process. The question is what that process costs the person inside it.

Manipur: when absence is not disappearance

Manipur brings another complication into the picture: displacement.

The state’s SIR draft contained 19.34 lakh electors out of 20.93 lakh on the existing roll. According to Chief Electoral Officer Arun Kumar Sinha, 1.58 lakh electors, or 7.58 per cent, had not submitted enumeration forms.

He said that voters whose names were missing could apply through Form 6 and that those whose records required verification would be given an opportunity to produce documents.

Sinha also stressed the procedural safeguard: “Nobody can remove anybody from the draft electoral roll unless he has been given an opportunity of hearing.”

But the operation has produced specific disputes over omissions.

In Churachandpur Assembly Constituency, 737 Meitei voters were reported missing from the draft electoral roll, according to the Churachandpur Meitei United Committee and The Sangai Express.

The newspaper reported that the remarks against omitted voters included “Missing”, “Absent” and “Dead”, as well as shifting of residence. The committee said many of the affected voters had submitted enumeration forms through nodal officers, but their names had not appeared in the draft.

The committee subsequently said the affected voters had begun filing Form 6 and that fresh enumeration forms would be submitted.

The significance of the episode extends beyond the number 737.

Manipur’s conflict has displaced people from their homes and complicated the relationship between physical location and electoral records. A voter may be living somewhere else without having abandoned their electoral connection to the place where they were registered.

The Churachandpur case does not by itself establish why every one of the 737 names was omitted. That is precisely why the verification process matters.

The question is not whether a safeguard exists on paper.

It is whether the safeguard can work effectively when the people being verified have already been displaced.

Nagaland’s 2005 problem

Nagaland brings the historical-record problem into sharper focus.

The state’s current SIR uses the previous intensive revision of 2005 as a reference point. The scale of the resulting mapping problem is striking.

According to Additional Chief Electoral Officer Awa Lorin, 9.33 per cent of electors fell into the “no-mapping” category and another 27.31 per cent into “mapping anomalies”. Together, those categories accounted for 36.64 per cent of electors covered during the exercise. Nearly five lakh voters were expected to receive notices over discrepancies in mapping their details with the 2005 roll.

The draft electoral roll published on September 20 contained nearly 1.59 lakh fewer names than the 13.57 lakh electors covered by the exercise. Electors whose names could not be mapped to the 2005 roll, or who could not establish themselves as the progeny of someone included in that roll, were to receive notices and produce prescribed documents.

This produces an obvious question.

How often does a 2005 baseline fail to map cleanly onto a 2026 voter?

A historical roll is useful evidence. It is not necessarily a complete description of the electorate two decades later.

For an elector who was a child in 2005, relevant evidence may belong to a parent. For someone who has moved between districts, the difficulty may be one of record linkage rather than eligibility. For a person whose name has changed spelling across documents, the database may register a discrepancy that a human official can resolve relatively easily.

The procedure exists to deal with these situations.

What deserves greater attention is how often the initial database classification changes after human scrutiny.

The Supreme Court has already asked the question

The distinction between a database flag and a final deletion is no longer only an administrative question.

It has reached the Supreme Court.

In September, while hearing petitions concerning Delhi’s SIR, the Court said notices appeared to have been issued “mechanically” to voters classified as doubtful and asked the Election Commission to examine whether additional filters were required. Justice Joymalya Bagchi said, “Refinement of the tool is necessary” and suggested that many of the notices appeared to be machine-generated.

The petition before the Court was filed by activists Anjali Bhardwaj and Amrita Johri. They sought disclosure of the names and reasons for notices issued to more than 33 lakh voters under categories including “logical discrepancy” and “unmapped”. They also sought information on the criteria and algorithmic parameters used to classify voters under those categories.

The broader question raised by the hearing was more basic: do voters receiving these notices know what they are being asked to establish?

The Election Commission has maintained that no voter will be deleted without due process and an opportunity for hearing.

The Court’s intervention puts another question into the debate: how transparent should the process be by which a voter becomes a case requiring verification?

That question becomes harder in the Northeast, where the historical reference itself can be complicated by displacement, migration and decades-old records.

The case for verification comes from the Northeast too

There is a serious argument for maintaining accurate electoral rolls, and the Northeast itself provides evidence for it.

Assam was handled separately from the current SIR phase because of its distinctive citizenship and electoral history, including the NRC process. It instead underwent a Special Revision involving house-to-house verification.

During that exercise, Booth Level Officers identified more than 4.78 lakh names for deletion because electors had died, 5.23 lakh because voters had shifted from their registered locations, and 53,619 as duplicate entries requiring correction. The final roll published in February 2026 contained 2,49,58,139 electors.

Those figures show why electoral rolls need periodic verification.

People die. People move. Duplicate entries occur. Records become outdated. In a region where migration and citizenship have been politically contentious for decades, there is a legitimate administrative interest in knowing whether the electoral roll accurately reflects eligible voters.

The issue, therefore, is not whether electoral rolls should be revised.

It is whether verification and deletion are being treated as the same administrative act.

The Assam experience is instructive precisely because house-to-house verification generated categories that required further scrutiny and subsequent processing. The value of an electoral database lies not only in its ability to flag a problem, but also in what happens after the flag is raised.

What the EC rift actually changes

This brings the story back to the 14 objections inside the Commission.

The significance is not simply that commissioners disagreed.

Nor is centralisation automatically suspect. A national digital architecture can make electoral administration faster and more consistent. In a country with hundreds of millions of voters, the alternative to centralised technology is not necessarily better human judgment. It can also mean slower, fragmented administration.

The real question is where the central system’s authority ends. A database can identify a mismatch. A BLO can locate the voter. An ERO can examine documents and hear the case. The institutional design matters because those stages are supposed to correct one another.

For the Northeast, the stakes are unusually clear.

In Manipur, 737 Meitei voters in Churachandpur were reported missing from the draft roll, with the local committee attributing the omissions to categories including absence, death and shifting of residence.

In Nagaland, 36.64 per cent of electors fell into either the no-mapping or mapping-anomaly categories during the SIR exercise.

For a first-time voter, the relevant historical record may belong to a parent.

For a voter whose name has changed spelling across documents, a database may see two different records where a local official can establish that they belong to the same person.

The central system can raise the question.

The human system has to answer it.

That is why the Election Commission’s internal disagreement over data access and control matters beyond Delhi. The issue is not whether technology belongs in election administration. It clearly does.

The issue is whether the voter can see, challenge and correct the decision produced by that technology.

The most useful numbers are therefore not only deletions.

They are the number of voters flagged, the number served notices, the number restored, the time taken to restore them, and the proportion of cases in which human verification changes the initial classification.

Those figures would allow the country to examine SIR as an administrative exercise rather than a political slogan.

The answer is not a softer electoral roll. It is a more transparent one.

An accurate electoral roll must prevent an ineligible person from voting. But it must also ensure that an eligible citizen is not pushed out because a decades-old record, a spelling variation or a central database failed to recognise them.

After 14 objections from within the Election Commission, the question is no longer simply whether the system can find discrepancies.

It is whether the institution can show, case by case and state by state, what happens when the discrepancy is the system’s, and publish the data that proves it.

Also Read: POCSO accused Santanu Kumar Nath secures interim bail