NUS study: Legal reforms shift public discourse on social issues, but institutional and societal conditions determine lasting change

The past decade has brought a wave of legal reforms around same-sex relationships across Asia: India struck down its colonial-era ban in 2018; Taiwan legalised same-sex marriage in 2019; Singapore repealed Section 377A, with the change taking effect in 2023; and Thailand’s marriage equality law came into effect in 2025.

These legal changes raise a fundamental question: does legal reform actually lead social change? And if so, how quickly, and under what conditions?

A new study led by NUS Communications and New Media (CNM) Associate Professor Kokil Jaidka suggests the answer is more nuanced than a simple “yes” or “no”. Analysing over 200,000 social media posts and survey responses across all four countries, the research reveals that while legal changes do reshape public discourse, the path to social acceptance depends heavily on the surrounding institutional and media environment.

Language shifts when the law changes, but not uniformly

Published in the journal Nature Human Behaviour, the study examined 65,969 Instagram and 136,255 Facebook posts before and after each country’s legal reform. In Singapore, Taiwan and Thailand, moral stereotyping and hostile language on Instagram declined in the months following reform. Users who were already engaged in these conversations began framing them with less reliance on language depicting same-sex relationships as immoral or a threat to tradition.

“This shows that legal reform does reshape the terrain of public expression, at least for those already paying attention,” said Assoc Prof Jaidka, who led the study with co-authors Ms Preetika Verma and Ms Yiting Chen from Carnegie Mellon University and University of Wisconsin-Madison respectively.

However, looking across different countries and platforms painted a more complex picture. The findings also show that changes in public discourse following legal reform are multifaceted and varied across contexts.

In India, where the researchers could track a longer period after decriminalisation, hostile language on Instagram increased even as moral stereotyping fell.

On Facebook, where news outlets generally reach broader audiences, changes were modest across all four countries. Engagement with LGBTQ+ topics did not consistently increase after legal reform, and, in some cases, appeared to decline after the initial period of public discussion.

Drawing on nationally representative Gallup survey data from 46,200 respondents, the researchers further discovered that perceptions of neighbourhood safety for same-sex couples initially declined following legal reform before recovering in some age groups and countries. Younger respondents in Singapore and India eventually reported more positive perceptions, but the initial dip serves as an important reminder that a change in law does not always immediately translate to lived experience.

What explains the variations? The study points to institutional context as a crucial moderating factor. In countries with greater press freedom and freedom of expression (Taiwan ranked highest on both measures), reductions in hostile online discourse after reform were steeper and more sustained. Where these conditions were weaker, the gains were smaller or more mixed.

“The law creates a moment of change, but the wider environment such as society and the media shapes whether that moment becomes lasting progress,” observed Assoc Prof Jaidka.

When AI tools mistake identity for hostility

The research also uncovered an unexpected challenge in measuring online discourse: the very tools designed to detect toxic language can inadvertently flag non-toxic content. When the researchers analysed Instagram captions using Google’s Perspective API, one of the most widely used toxicity classifiers, they found that identity terms such as “queer”, “gay”, and “lgbt” were among the words most strongly correlated with high toxicity scores, despite not being slurs.

As individuals from marginalised communities are disproportionately targeted by online harassment, the words describing their identities appear frequently in abusive content. The algorithms then associate the identity terms themselves with toxicity, potentially suppressing ordinary conversations by the very communities they aim to protect.

The researchers explained how a toxicity score is a relative signal shaped by biases in training data, not an objective verdict on whether speech is harmful, adding how this revealed the need to use these tools with awareness of their limitations.

What comes after legal change

The findings have practical implications for multiple stakeholders. For online platforms, the period immediately following legal reform appears particularly consequential, as communities recalibrate and some voices push back. For media organisations and community groups, the research underscores that legal milestones mark a pivot toward sustained engagement around social attitudes, public discourse and acceptance, rather than a closing chapter.

Assoc Prof Jaidka said, “The research offers reasons for both encouragement – as legal reform does measurably shift discourse – and continued investment in the institutional and social conditions that translate legal changes into lived experience across the region.”