Dispatches on Tech and Democracy: Five Things We Can Learn from India's 2024 Elections

Amber Sinha is a Tech Policy Press fellow.

The national elections in India, held over a long and grueling eight weeks of summer, finally concluded last week. Over the last two months, these dispatches have been bringing you regular discussions on the use of technology, data, and AI in India’s elections. Contrary to expectations of a landslide victory for the incumbent Narendra Modi government, his party was able to come back to power only with the support of coalition partners. Over the last week, substantial media coverage has focused on explaining this slight reversal of fortune for the BJP in India.

Providing any such analysis which explains the electoral vicissitudes in a complex, diverse and large country like India is inevitably an exercise in oversimplification. Rather than try to deduce the whys of the election results, the dispatch will look at the broad trends, the insights they may provide, and what they might mean for future elections in India and the rest of the world.

1. Digital infrastructure is critical for political parties.

The 2024 elections were a timely reminder of the extensive digital infrastructure and tools employed by political actors in India for the purposes of electioneering. I use the word ‘infrastructure’ broadly to include a wide range of digital tools and networks built by political parties, including the use of social networking sites, dedicated mobile phone apps, WhatsApp groups, arrangements for data collection and access, and the use of AI-based synthetic content for personalized calling.

India has roughly 820 million Internet users, staggering in its scale and second only to China, but still accounting for less than 60% of its population. Smartphone users only comprise 46% of the population. This means that the potential for the use of digital technologies will only grow dramatically in the coming years as mobile phone and Internet usage expands to rural and remote parts of the country. What this suggests is that as extensive as the organized digital infrastructure of political campaigns may be currently, it is only a fraction of what we may witness over the next decade.

Related Reading:

• Dispatches on Tech and Democracy: India’s 2024 Elections #4
• Dispatches on Tech and Democracy: India’s 2024 Elections #5
• Dispatches on Tech and Democracy: India’s 2024 Elections #6

The other key insight to draw from the state of digital electioneering infrastructure is the relative regulatory vacuum in which it exists and operates. The first factor is the nascent nature of regulations around difficult governance issues like speech and content moderation of WhatsApp groups or emerging issues like deep-fakes, generative AI, and synthetic content. The second factor is that despite robust examples of international regulations, the Indian state has chosen to allow various activities to go comparatively unregulated. These include the absence of campaign financing laws, which would allow for stricter audits of money trails, failure to frame regulations around online political advertising, and poor or virtually non-existent data protection regulations.

These gaps allow for political campaigning, advertising, and targeting on digital platforms to skirt existing laws at the disposal of India’s election management body, the Election Commission of India, and other regulatory actors. These digital experiments have implications not only for future elections in India but also offer novel strategies for sophisticated, well-funded, and organized political actors in other mobile-first Majority World economies with a powerful executive and low state capacity.

2. Platforms like YouTube provided space for alternative media sources and counter-narratives.

The last decade and a half has witnessed a tussle between traditional news media operators in print and television, and the consumption of user-generated content on digital platforms and social networking sites. Even in this short time period, this tension has played out in many different ways, with actors on digital platforms resorting to innovative strategies to respond to or compete with traditional news media.

The last decade has also seen a largely pliant and establishment-friendly print and television news media in India. This has to do with market factors where the ruling party and the government are some of the biggest contributors to advertising revenue, the use of state’s enforcement agencies such as the tax authorities and enforcement directorate against some of the voices critical of the government, and large scale investments or acquisitions of news media, particularly television channels by businesses close to politicians in power. The closeness of news media to the BJP-led government is not merely an observation of journalists, opposition politicians, or outside studies, but increasingly a popular perception.

In this election, we saw this perception translating into significantly more eyeballs for political analysis and content on online platforms, particularly YouTube channels. This included both journalists, formerly on traditional media, setting up channels on Youtube, as well as online-first influencers whose political analyses attracted huge followings. The prominence of these online voices is said to have countered the increased coverage Prime Minister Modi and his party enjoyed in print and television media and offered an alternative narrative.

This tussle between mainstream and alternative media promises to be the subject of regulatory debate in the government’s digital policymaking. Last year, draft legislation was circulated by the Ministry of Information and Broadcasting which seeks to bring over-the-top (OTT) services within the purview of broadcasting regulations. The bill proposes content evaluation committees that would certify content that streaming services host.

3. Existing regulations and codes of ethics for online political advertising in India are inadequate.

In a previous dispatch, I argued that ECI’s Voluntary Code of Ethics for regulating political advertising online was inadequate. Neither the ECI nor the Representation of Peoples Act specifies a definition of the term “political advertising.” Thus, the various Internet platforms were largely left to determine how they would define and govern political advertising.

As I describe in more detail here, although political parties and candidates were required to disclose their expenditures on social media advertisements, there has still been no attempt to regulate ad spending by loosely connected supporters who indirectly contribute to campaign funding through coordinated advertising. Some platforms have made efforts to introduce accountability in this unregulated space, such as requiring disclaimers on paid advertising, implementing take-down procedures for non-compliance, and creating a public repository for easy access to advertisements and expenditures. However, these measures are still inadequate in identifying all types of political content and actors and are only partially enforced.

In a report focussing on large spending by shadow advertising networks on Facebook promoting hate speech, communal violence, and disinformation in India, The London Story, along with Eko and Indian Civil Watch International, provide a clear path for the ECI to regulate online political advertising. It relies on laws and regulations, and, most importantly, progressive rules set forth by the Supreme Court of India that regulate political and third-party advertisements on political issues during specific periods of elections. These rules emerge from orders of the court in Ministry of Broadcasting v. Gemini TV Pvt. Ltdin 2004, which led the ECI to require that all political parties, along with contesting candidates intending to broadcast advertisements on television channels or cable networks, must seek pre-certification. In 2013, the ECI expressly extended the scope of this obligation to advertisements on social media platforms.

Despite this, the above report highlighted that it is “highly unclear if Meta is seeking any kind of pre-certification of ads, especially those put out by coordinated networks. The rules provided by the Supreme Court provide a clear mandate and framework for ECI to effectively regulate all paid political advertising during elections but the lack of transparency impedes our ability to see the extent to which they are being enforced. This provides us with some clear and strategic points of intervention for advocacy — the need for more commitment to trust and safety from platforms to address harms on their platforms, greater transparency from platforms on paid advertising and whether they fall within their definition of political ads, capacity building and more resources for election management bodies, and investments into tools, standards, and frameworks which can help Election Management Bodies (EMBs) evaluate paid political content.

4. Public access to sensitive voter data creates risks for misuse by political parties and others.

The surprising results of this election also raise questions about the issues that were important to voters and how the political campaigns responded to them. Fundamentally, election campaigns need to predict the motivations of the voters, which voters will turn up to the polls, and how likely they will support one candidate over the others They need information on what voters prioritize, how best to reach them, and perhaps more specific questions such as what issues are important to them and are they persuadable In Hacking the Electorate, political scientist Eitan Hersh points out that while these perceptions of a campaign are based both on “gut feeling” and “hard data,” the campaign’s perception about a voter’s attributes is likely to be different from a voter’s own self-perception.

The economic theory of democracy entails that “from the basic economic nature of becoming informed emerges the necessity of selection among data.” This immediately leads to questions about how data must be selected, and it determines what type of information is instrumental in making decisions. This plays out in both ways—how information and perceptions affect the ways that voters decide and how polls, data, and other sources of information affect how campaigns choose to represent voter interests.

The most commonly used source of data in India for political parties has always been government electoral data. The ECI maintains electoral rolls for each district and, more importantly, for each polling booth. These are public databases that include details about voter names, age, gender, father or husband’s name, address-related information, the photo identity card number, and, in some cases, also photographs. This can easily lead to inferring information about caste and religion, and profiling of a number of voters residing at a particular address and their demographic makeup. It is worth noting that this is all legally available public data, which can be mined by anyone. Further, the absence of an implemented data protection law in India means that there are no regulations that govern how this extremely sensitive personal data is collected, masked, shared, published, or sold.

The Registration of Electoral Rules requires the registration officer to publish the roll with a list of amendments at his office for inspection and public information. The Election Commission may also direct the registration officer to send two copies of the electoral roll to every political party for which a symbol has exclusively been reserved. The electoral roll of a constituency is a public document given that the roll is published and can be circulated in the direction of the ECI. In 1998, the Election Commission decided to digitize the electoral databases. The publication of electoral rolls was not a significant privacy risk when it was distributed in hard copies. The privacy risk emerged only after the digitization and online publication of the electoral database.

Unlike the US, there is no organized system of registered party affiliations in India. The range of options in a multi-party system like India is also much larger. However, it is possible to look at the demographic makeup of a constituency and correlate it against past voting records to arrive at broad conclusions. Where there is more data available on party affiliations, campaigns may likely focus more on mobilizing supporters and trying to persuade undecided voters, rather than engaging with opposing voters. Traditionally, the focus of election campaigns has been more on geographic-level targeting based on past election returns. However, with more granular data available, it is becoming easier to target campaigns at individual levels.

The availability of a centralized and searchable database of voters along with their age, would allow the ECI to identify wards or constituencies, which, for instance, have a high population of voters above the age of sixty-five. This would help them set up polling booths at closer locations with special amenities. However, the same database at the hands of political parties can be used to search for the density of members of a particular community in a ward or constituency based on the name, age, and sex of the voters, and frame campaigns to manipulate voters. Despite some of the standardization issues, it is easy to write code that will scrape data from electoral rolls available online and sort them into databases, which can then be used for analysis.

An investigation by The Wire published immediately after the elections found a network of data brokers and several booth management mobile apps that supply voter data to political parties, and that they had been used by at least 4,000 candidates in the last few years across the country in local corporations, state assembly, and the 2024 national election.

The Indian example presents an important reminder for better regulation of personal data used by political actors and the need for data protection authorities and EMBs to have better coordination and clarity on the transparency and privacy objectives they fulfill.

5. While generative AI was not a major issue, more subtle AI applications remain a concern.

In a conversation with Mark Scott of Politico, before the national elections commenced, I argued that fixation with generative AI — the most advanced uses of the tech powering deep fakes — missed the most likely way for AI to affect this year’s global election cycle. Two months later, there is an overwhelming consensus that the Indian elections were not the generative AI disaster everyone thought they would be.

However, as my colleague Vandinika Shukla has argued here, the elections were instructive on the ways in which synthetic content may be deployed in future elections and the kind of impact it might have. The use of AI in personalizing voice calls by political campaigns or live voice translations of speeches in regional languages were clear examples of the use of technology to make political messaging more effective.

More importantly, the use of machine learning in more prosaic forms by social networking sites for powering their recommendation engines played a more important role than the eye-catching use of generative AI and synthetic content. As mentioned above, the consumption of news and information on platforms like YouTube played an important role in this election, and the documented tendency of YouTube Video recommendations to lead to more extremist content assumes even greater importance. Similarly, investigations during this election highlighted hate speech and misinformation on Meta’s Facebook platform, whose machine learning-fueled algorithms have been found to steer users to increasingly extreme viewpoints.

Outside of AI-driven recommendation engines on online platforms, the use of AI to analyze demographic, household, and personal data drawn from a variety of sources, make inferences about the political motivations of citizens, and help campaigns manipulate them remains another big factor. The potential for this form of AI use to grow further along with the digital electioneering infrastructure is significant.