An engineer, a historian and a physician. Conventions and stereotypes place the three authors of this essay at different points across the STEM (science, technology, engineering and mathematics) and humanities divide: designing systems, interpreting the past, and using biology and evidence to serve a patient. 

Yet, we have much in common: All of us need to build trust and relationships, communicate well, work in teams and understand human nature to do our jobs well.

Our training gave us different bodies of knowledge and different habits of mind. It also gave each of us a discipline rooted in critical thinking: a way to define a problem, decide what counts as evidence, test a claim and recognise error.

As our careers developed, the boundaries between our fields of work became less obvious. The engineer had to understand institutions, incentives and public acceptance. The historian had to work through patterns, causal claims and incomplete records. The physician had to move between data from populations and decisions about one person. By the time we reached leadership roles, each of us had learnt skills associated with the other side of the divide. 

In essence, we each had to close the divide between STEM and the humanities. Crossing the divide did not make our primary disciplines any less important. Singapore’s development depended on disciplinary expertise. We will continue to need depth in each field.

But the problems we face today are increasing in complexity, scale and impact. Climate adaptation requires engineering models and consideration of questions of cost, behaviour and fairness. Urban planning increasingly requires the assessment of transport data, history, place and community. Digital regulation needs depth in understanding software architecture and markets. Tackling these problems requires people who are able to draw on different bodies of knowledge, not just those from their field.

Our personal journeys have led us to conclude that Singapore will need to address the need for multidisciplinary skills development at the level of the individual, the team and the nation. This need is not new, and it was not created by artificial intelligence. But we believe AI can accelerate the closing of the STEM-humanities divide.

Increasingly, natural language is becoming an interface through which people can call upon capabilities beyond their formal training, with AI agents handling much of the technical execution. This does not require everyone to become a programmer. But effective use of a large language model (LLM) requires more than access to a model. 

What remains essential is the ability to think computationally: to decompose a problem, define its logic, sequence tasks, specify constraints and anticipate how a system may interpret an instruction. The user becomes a task architect: dividing the work into parts, delegating some of them to AI agents and shaping what comes back into a coherent whole. 

Users need enough understanding to know what the system is doing, enough subject knowledge to judge its output, and enough knowledge of language and context to frame a task. They must also decide when the output can be trusted, who bears responsibility for it and whom it may harm.

The challenge is that our disciplines have become a source of identity. Students see themselves as technical or verbal, quantitative or creative. Institutions reinforce the categories through admissions processes and faculty appointments. Yet the STEM-humanities divide is smaller than our institutions and labels suggest. The disciplines are distinct, but the skills they develop overlap, and the problems we face often require several of them at once.

Hence, we need a multidisciplinary mindset, rather than to label ourselves as one side of the STEM-humanities divide or the other.

We don’t mean to suggest that we should blend disciplines until their methods lose definition, nor that everyone must become an expert in several fields. Rather, we all need grounding in a discipline and to achieve working fluency in others, which AI now puts within reach, so that several disciplines can be brought to bear on one problem. 

Building a multidisciplinary mindset among individuals is just the first step. We have to close the STEM-humanities divide as individuals, teams and institutions.

For a multidisciplinary team to work well, its members must understand one another. Engineers must know enough about social inquiry to recognise a behavioural assumption. The policymaker needs to know enough about models to question a metric. A lawyer should understand where an AI system is deterministic and where it is probabilistic. As AI becomes an integral part of research, the historian will need to complement traditional skills with new habits using language to guide digital tools. 

This extends across many jobs and industries. To be promoted from technical operations to business management, to take a content creation side gig and turn it into one’s main source of income as a media platform, to start a new cafe after learning to cook as an employee; all these and more require skills from across a range of disciplines.

Closing the STEM-humanities divide across a whole society is a large task. But Singapore has taken on a comparable challenge before, in our bilingual policy. English gives us access to a working language shared across communities and connected to the world, and our mother tongue languages connect us to culture, history and family.

We built policy, teaching and social expectations around the value of operating in more than one linguistic world. We now need a second bilingualism to close this divide: fluency in the language of our own discipline and the language of AI.

Tertiary education institutions will need to respond to prepare their students for a lifetime of acquiring skills, based on a foundation of disciplinary expertise, matched with a fluency in AI.

Institutions can act on this without weakening disciplines. Students should still learn mathematics, science, computing, literature, history, languages and the arts through the standards of those specialised fields. And they should also encounter problems that require several disciplines, taught by people who can make the differences in method explicit. Students need exposure to more than one evidential tradition to learn that rigour need not always require a measurement, and that measurements need
interpretation.

This is multidisciplinary learning. A climate problem can be examined through physics, economics, history and political judgment. A language model can be studied through mathematics, computer science, linguistics, philosophy and law.

AI gives us a place to begin because it challenges the categories we operate in. Ask how a language model works and the answer begins with mathematics. Ask whether it understands and the answer moves into philosophy and linguistics. Ask whether to deploy it in a hospital, court or classroom and the answer requires professional knowledge, law, ethics and public judgment. Ethical reasoning cannot replace technical knowledge, and one cannot judge a system’s use without knowing what it can do, where it fails and how it is deployed. Technical knowledge cannot replace ethical reasoning either. 

The problems we face are already multidisciplinary. Our learning should prepare us for them.

What would this entail? First, assess students on whether they can frame a problem, choose evidence from more than one field, state the limits of each method and build a decision from them. If assessments reward work within one subject boundary, students will treat cross-disciplinary fluency as enrichment. 

Second, plan for encounters across the curriculum and time for teachers from different fields to design them. It does not mean a new omnibus subject.

This is not an argument to teach all students everything about all subjects. We believe that in tertiary education there is still a need to be grounded in a domain, and to develop mastery of that subject. 

Nor are we arguing that all tertiary students should be fully multidisciplinary before employment. Learners develop through repeated work near the edge of their current ability. A student may debug a program, reconstruct an argument from sources, or revise a diagnosis after new findings. 

In each case, productive struggle turns rules into competence and, with experience, competence into judgment. It takes time for the skills to mature and for the individual to grow. But the foundations for that journey should begin within our tertiary institutions.

Singapore’s advantages will be people and institutions that are able to make several forms of expertise work together: preserving disciplinary depth while multiplying what those disciplines can achieve in concert.

We built one form of bilingualism; we can build another. This will require changes in curriculum, assessment, professional training and leadership. The engineer, the historian and the physician should still bring different knowledge to the table and have the confidence that each can understand what the others see, so that they ask better questions and reach better decisions together.