AI Era Puts Singapore’s Hasty Reading Problem Under a Sharper Lens

Channel News Asia documents a reading problem that is no longer just about books: Singapore students still lead global rankings, but teachers and researchers are seeing more young people skim, rush and miss meaning as digital habits and AI shortcuts reshape how they process text.

The warning lands inside the 2025 OECD Programme for International Student Assessment, released on September 8.

Singapore’s 15-year-olds kept the top reading position, but the score moved down to 535 in 2025 after standing at 543 in 2022.

The broader PISA pool covered 760,000 students across 91 countries, where the average reading result also weakened, dropping to 461 from 476.

The OECD framed the decline as especially important in an AI age because reading now underpins the ability to evaluate information, connect multiple sources and think critically about what has been read.

Its results also showed that “hasty readers” — students who move through passages quickly and give fast but wrong answers — nearly doubled between 2018 and 2025.

Inside classrooms, the statistical shift looks more ordinary and more troubling.

Secondary mathematics teacher Julian Teo has seen students skip words in problem sums or begin solving before they finish reading.

One example he showed was a 146-word question about the cost of driving between four cities, a length that was enough to expose how difficult “chunks of words” have become for some learners.

The issue is not limited to mathematics.

An upper primary English teacher using the name Jennifer described 11- and 12-year-olds who appear to read quickly but are often moving from paragraph to paragraph out of boredom.

Some students told her they skip book chapters to reach the conclusion and move on to the next entertaining thing, turning the arc of a story into another item to clear.

Older students show a similar pattern.

Kelvin Hong, who founded The Economics Tutor and Ace GP and has taught Economics and General Paper for more than 25 years, has found that more junior college students struggle to explain a writer’s overall argument, recognise nuance or connect ideas.

Long articles and complex passages increasingly push them toward keyword hunting rather than following a line of reasoning from beginning to end.

The source points to several layers behind the shift.

The Turkey classroom study covering 1,000 students during the 2023 to 2024 academic year produced a clear contrast: unrestricted AI access was associated with weaker closed-book test performance than no AI access or guarded use.

A separate Singapore-Australia study covering 583 people between 13 and 25 produced the same attention warning in a different form: 68 percent had trouble focusing, one-third characterised compulsive scrolling as addictive, and 52 percent linked classroom distraction to social media.

Adult habits reinforce the same environment.

A 2024 National Library Board study found that only 28 percent of Singaporean adults read books weekly.

When news, social media articles and websites were included, the reading share rose to 89 percent, showing that short-form and screen-based reading are far more common than sustained book reading.

Among adults who did not read books more often, 62 percent preferred other activities and 38 percent became bored after reading for a while.

For learning specialists, the concern is not that technology exists but that it can crowd out the mental work that literacy requires.

Annabel Chen, a clinical neuropsychologist at Nanyang Technological University, described digital environments that fragment attention through notifications, task-switching and competing information.

Search engines and generative AI can also make it easier to offload retrieval, integration, formulation and evaluation — useful in moderation, but risky when they replace practice in building knowledge.

That risk becomes sharper when AI produces fluent explanations.

A polished summary can create the feeling that a student understands a topic even when the argument has not been checked, reconstructed or explained independently.

In that setting, quick access to information does not automatically become comprehension.

The literacy definition is widening at the same time.

Associate Professor Loh Chin Ee of the National Institute of Education described a literate person as someone able to process both print and digital texts, including images, video and graphics across multiple sources.

That means judging whether information is reliable, deciding whether to act on it, and recognising how convincing AI-generated material can still be inaccurate.

The labour market gives the classroom problem a longer tail.

Foo See Yang of recruitment provider PERSOL APAC linked AI literacy to future job opportunities, arguing that employers will look beyond tool use to judgment and critical thinking.

His practical sequence is to attempt a task first, form an initial view, then use AI for feedback, refinement or another perspective so the user stays engaged in the thinking process.

Singapore’s Ministry of Education is trying to draw that line by age.

Primary 1 to 3 learning remains hands-on and print-first, with limited use of technology and AI.

Upper primary students use AI under teacher supervision, while secondary students move gradually toward more independent use through structured tasks.

The ministry also emphasised qualities that AI cannot replace, including critical thinking, self-management, values and responsible choice.

Hougang Secondary School shows how that can look in practice.

Its STEM Applied Learning Programme asks students to evaluate evidence, make connections and draw on different sources, with AI used where relevant to generate and test solutions.

Infocomm technology department head Javen Gwee framed the objective as judgment: students should know when AI helps their reasoning, when its output needs challenge and when the tool should be set aside.

The answer, then, is not a retreat from digital tools.

Reading time in schools, meaningful long-form texts at home, careful modelling by parents and structured AI use all serve the same purpose: preserving the patience to stay with an argument before asking software to shorten it.

If students lose that habit, the problem will not be slower exam answers.

It will be a weaker ability to decide what is true when information arrives instantly.