Google’s search engine handles the vast majority of online searches worldwide. By one estimate, it fields 6.3 million queries every second.
Because of the search engine’s enormous scale, its outputs can have outsized effects. And, while Google’s search results are shaped by ostensibly neutral rules and processes, research has shown these algorithms often produce biased results.
This problem of algorithmic bias is again being highlighted by recent escalating tensions between India and Bangladesh and cases of violence against Bangladeshi citizens in India and violence against Hindus in Bangladesh. A pro-Indian misinformation and disinformation campaign is exploiting this algorithmic bias to further its agenda – an agenda that has been described as Islamophobic and alarmist.
This kind of misinformation has been implicated in several riots and violent incidents in Bangladesh.
All of this serves as an important reminder of the power Google’s search engine has in shaping public perceptions of any event – and its vulnerability to being exploited. It’s also an important reminder to anyone who uses Google’s search engine to engage critically with the results it dishes up, rather than accepting them at face value.
What is algorithmic bias?
The algorithms that power Google’s search engine are trained on massive amounts of data. This data is gathered by computer bots which crawl billions of pages on the Internet and automatically analyse their content and quality. This information is stored in a large database, which Google’s search engine relies on to serve up relevant results whenever it receives a query.
But this process doesn’t capture every website on the Internet. It is also governed by predetermined rules about what is high quality and what is low quality, and reflects existing biases in data. For example, even though only 16% of the world’s population speaks English, it accounts for 55% of all written content online.
This means the reality of life on the ground in non-English speaking countries is often not reflected in Google search results. This is especially true for those countries located in the Global South.
This lack of representation perpetuates real-world biases. It can also hinder a nuanced public understanding of global issues.
What’s happening between Bangladesh and India?
Relations between Muslim-majority Bangladesh and neighbouring India, which is currently led by the Hindu nationalist BJP government, have deteriorated recently.
In August last year, youth-led anti-government protests erupted in Bangladesh.
These protests resulted in the downfall of prime minister Sheikh Hasina’s long-lasting autocratic regime, which had been supported by the Indian government.
An interim government filled the void. But certain Indian media outlets have leveraged sensitive issues such as Hindu minority rights to undermine its legitimacy.
In November, Bangladeshi authorities arrested Hindu leader Chinmoy Krishna Das on sedition charges over allegations he had disrespected the Bangladeshi flag. This triggered violent clashes between his supporters and police. These clashes resulted in the death of a Muslim lawyer.
Hindu activists also attacked a Bangladeshi consulate in India.
There have also been verified instances of mob violence against Hindus in Bangladesh. However, the Bangladeshi government claims these incidents are politically motivated rather than communal attacks.
The unrest intensified earlier this month, with thousands of protestors destroying the family home of deposed prime minister Sheikh Hasina in the Bangladeshi capital, Dhaka.
Boosting a disinformation campaign
A disinformation campaign based in India has exaggerated some cases of religious violence against Hindus in Bangladesh.
This campaign has been boosted by Google’s algorithmic bias.
For example, an analysis by the Tech Global Institute of Google search results about Chinmoy Krishna Das’s arrest between November 25 and December 20 last year found a “consistent pattern of bias”.
Specifically, Indian news outlets – including Hindu ultranationalist news outlets – “disproportionately” dominated the top search results. This overshadowed
factual reporting from credible Bangladeshi media outlets […] despite the search originating from within Bangladesh, the country where the incident originally occurred.
This bias was also evident in search queries coming from overseas. For example, roughly 90% of the top results about Chinmoy Krishna Das were from Indian outlets when searched from Australia and the United States. Bangladeshi news outlets featured on the thirteenth and fourteenth pages of results.
Indian news outlets – unlike their Bangladeshi counterparts – produce a substantial amount of content in English. They also employ more advanced search engine optimisation – or SEO – techniques, such as using effective keywords and sensationalist headlines. This gives them an advantage in Google search results compared to their Bangladeshi counterparts.
Another investigation by Bangladeshi fact-checking outlet Rumor Scanner in December 2024 found 72% of social media accounts spreading fake and misinformation are located in India.
The Conversation asked Google a series of questions about its search engine. It did not receive a response.
An illustrative case of a global problem
Bangladesh is an illustrative case of the global problem of algorithmic bias. It highlights how search engines can be exploited to promote disinformation and misinformation and powerfully shape people’s perceptions about what’s happening in the world.
It also highlights how everybody should think critically about the information they find online about the current situation in Bangladesh. Or about any news event, for that matter.
The case also reinforces the urgent need for policymakers, tech companies and governments to work together to effectively address algorithmic bias. This is especially urgent in the Global South, where marginal voices remain silenced.
Abdul Aziz, Lecturer in Media and Communication Studies, School of Arts and Social Sciences, Monash University
This article is republished from The Conversation under a Creative Commons license. Read the original article.
