Chief Executive John Lee Ka-chiu’s vision of an “AI city brain” is bold and timely. An integrated system linking data across transport, public works, environmental monitoring and emergency response could help Hong Kong manage risks more proactively and coordinate services more effectively.

The initiative is highly commendable. It also raises a critical question: can a city brain truly understand a city if it does not understand how its people communicate?

The truth is that communication is rarely monolingual in Hong Kong. People like myself move between Cantonese, Mandarin, English and other linguistic resources every day, often within a single conversation. This is not a problem to be solved. On the contrary, it is one of Hong Kong’s greatest strengths. The city’s advantage lies in its biliterate and trilingual population.

When policymakers discuss multilingualism, they often focus on how many languages people can speak. An equally important question is how people move between them. A multilingual person possesses multiple languages. A translanguaging person draws on them dynamically according to context, audience and purpose. Language use is not like a fixed set of separate mental compartments. It is a flexible system that allows people to convey and negotiate meaning, relationships and identity in real time.

The distinction is not merely semantic. A system built around multilingualism might classify a caller as being “Cantonese-speaking” or “English-speaking” and route the call accordingly. A system built around translanguaging would recognise the same caller might move between languages within a single sentence and that the movement itself could carry information. The first approach treats language as a fixed attribute, the second one treats it as a dynamic resource.

This matters because the proposed AI city brain will interact with human behaviour, not just infrastructure. The government should therefore include linguists and cognitive scientists in its development. Language is not simply another data set. It is one of the primary ways through which people experience public services, report emergencies and communicate their needs.

To see why this matters in practice, consider a realistic emergency scenario. During a typhoon, a hotel receptionist calls 999. She begins in Cantonese, switches briefly to English to describe technical details about the building, then returns to Cantonese to reassure frightened guests nearby. The information conveyed is not merely linguistic. The language shifts themselves reveal context, urgency and social dynamics.

An AI system designed around rigid monolingual categories might capture the words but miss part of the meaning that could help save lives at a critical moment. A language-aware system should not replace the human operator. It should preserve the linguistic context of the call, flag possible misunderstandings and suggest clarification questions. This is not about automating empathy. It is about giving human decision-makers better information so they can exercise judgment more effectively.

Emergency management is particularly relevant because the first phase of the AI city brain focuses on the Security Bureau’s Emergency Monitoring and Support Centre. In crises, communication can become more fluid rather than less. Research suggests people do not necessarily choose languages deliberately under stress. They draw on whatever resources are most accessible at the moment.

This does not mean artificial intelligence (AI) should infer people’s intentions from every language shift. Rather, it should recognise when multilingual communication creates uncertainty, preserve the original wording and encourage clarification rather than premature conclusions. In emergency response, recognising uncertainty can be just as important as recognising meaning.

This is especially important for vulnerable populations. Public systems are not designed only for the majority who can navigate them easily. They must also serve those who struggle most when circumstances become difficult. No one should be left behind.

The same logic extends beyond emergency response. Consider Macau’s gaming sector. A dealer might interact with a Mandarin-speaking visitor, an English-speaking tourist and a Cantonese-speaking local within a short period. The skill involved is not simply speaking three languages. It is moving between them smoothly while managing expectations, relationships and cultural cues.

As AI systems become more involved in customer support, public services and decision-making, they will increasingly encounter similar patterns across the Greater Bay Area. Should these systems merely classify people according to language categories, or should they recognise how language is actually used in real life?

Hong Kong is uniquely placed to address this challenge. Our competitive advantage is not simply that many residents know more than one language. It is that people routinely connect linguistic, cultural and professional worlds. Cantonese anchors our everyday life, English connects the city globally and Mandarin links Hong Kong to national development. Much of the city’s success comes from moving between these worlds rather than remaining within one of them.

An effective AI city brain should reflect that reality. A city brain that understands multilingual communication is more likely to make sound decisions because it is built around how people actually live, work and interact.

Hong Kong’s future competitiveness might depend not on speaking more languages but on understanding the value of moving between them. If our AI city brain is intended to serve and understand the city better, it should also understand the people who make the city what it is.