Religious Rehabilitation Group Exposes How AI Algorithms May Fuel Radicalisation

A Singapore-based research group has published findings that recommendation algorithms used by major technology platforms may inadvertently accelerate the process of online radicalisation, raising fresh questions about the economic liabilities facing social media companies and their investors.

Research Reveals Algorithm-Driven Pathway to Extremism

The Religious Rehabilitation Group, established in Singapore in 2002, conducted a study examining how artificial intelligence systems curate and recommend content to users. The research focused specifically on how seemingly innocuous starting points — such as cooking videos or fitness content — can gradually lead users toward increasingly extreme material through algorithmic suggestions.

The group presented its findings at a public forum, arguing that recommendation engines prioritise engagement metrics above all else, creating what researchers described as a dangerous funnel effect. When users watch one cooking video, the algorithm suggests another. Over time, these recommendations can shift toward content with ideological undertones, and eventually toward material promoting radical views.

The Salad Bar Comparison

Researchers used an accessible analogy to explain the phenomenon: just as a salad bar allows customers to gradually add increasingly spicy ingredients until the palate is overwhelmed, recommendation algorithms slowly introduce users to more extreme content. The study noted that this process often happens without users consciously seeking out radical material.

The Religious Rehabilitation Group has historically focused on rehabilitating individuals detained under Singapore's Internal Security Act. Their expertise in understanding how people become radicalised gave the study particular weight among policymakers attending the forum.

Regulatory Pressure Mounts on Tech Firms

The findings arrive at a time when technology companies already face mounting regulatory scrutiny across multiple jurisdictions. Governments in the United States, the European Union, and Asia have begun examining how algorithmic recommendation systems affect public discourse, mental health, and security.

Shanmugam, Singapore's Minister for Home Affairs and Law, has previously addressed the intersection of technology and national security. While he was not the primary presenter at this particular forum, his ministry oversees the Internal Security Department that works alongside organisations like the Religious Rehabilitation Group.

Liability Concerns for Technology Companies

The economic implications extend beyond reputational damage. Legal experts suggest that if regulators determine algorithmic recommendation systems directly contribute to radicalisation, technology companies could face significant liability. Class-action lawsuits, regulatory fines, and mandatory changes to platform architecture could follow.

Investors have already begun factoring these risks into valuations. Several major asset managers have issued statements indicating they are examining the algorithmic governance practices of their technology holdings more closely.

Market Reactions and Industry Response

Shares in major technology companies have shown increased volatility in recent months as regulatory concerns mount. The Religious Rehabilitation Group's findings add another layer of complexity for executives already navigating content moderation debates.

Industry representatives have pushed back against framing algorithms as inherently harmful. They argue that recommendation systems serve primarily to connect users with relevant content and that isolating algorithmic curation as a cause of radicalisation oversimplifies a complex phenomenon.

The Engagement Paradox

At the heart of the debate lies a fundamental tension in the technology business model. Platforms earn revenue through user engagement, and recommendation algorithms are designed to maximise the time users spend on services. Critics argue this creates a structural incentive to serve increasingly extreme content because such material tends to generate strong emotional responses and prolonged viewing sessions.

The Religious Rehabilitation Group's research suggests this engagement paradox may have unintended consequences that extend into national security territory. Their findings indicate that users who never sought out extremist content can still be gradually exposed to radical material through algorithmic pathways.

What Comes Next

Singapore's Ministry of Home Affairs is expected to review the Religious Rehabilitation Group's findings as part of ongoing assessments of national security threats in the digital age. The research could inform future policy recommendations regarding technology platform oversight in the city-state.

Internationally, the study adds to a growing body of evidence that regulators and investors are likely to use when evaluating technology companies. The European Union's Digital Services Act already imposes certain algorithmic transparency requirements, and similar frameworks are under consideration in other markets.

Technology companies are expected to face questions about their recommendation algorithms at upcoming earnings calls and regulatory hearings. How firms respond to these concerns — whether through voluntary changes, regulatory compliance, or continued resistance — will likely influence both their market valuations and their long-term regulatory exposure.