AI ethics clash with social work codes in Hong Kong and Singapore

Whose ethics govern when artificial intelligence enters the counselling room or the social worker’s caseload? A new study argues that in Hong Kong and Singapore, the answer is far from straightforward. Researchers examining the professional codes of ethics that govern social workers and counsellors in these two Asian jurisdictions have found that the rulebooks guiding these professions are strikingly silent on artificial intelligence, and that this silence is compounded by deeper cultural and structural mismatches between what the codes assume and how practitioners actually work.

The study, published in the journal AI & Society, was conducted by Gerard Siew Keong Chung and Yong Hao Ng of the Department of Social Work at the National University of Singapore, Pak-Hang Wong of the Centre for Applied Ethics at Hong Kong Baptist University, and Johnson Chun-Sing Cheung of the Department of Social Work and Social Administration at the University of Hong Kong. Through a systematic textual analysis of four professional codes of ethics—the codes of the Social Workers Registration Board of Hong Kong, the Hong Kong Professional Counselling Association, the Singapore Association of Social Workers, and the Singapore Association of Counselling—the team set out to assess whether these documents are adequate for governing practice in an era defined by artificial intelligence and what the authors call digital hybridity, the seamless blending of online and offline modes of professional work.

What they found is not merely a missing section on chatbots. The researchers identify three interconnected and compounding gaps. The first is a content gap: the codes, several of which were written well before the current wave of generative artificial intelligence, say nothing about artificial intelligence, algorithmic decision support, or generative technologies. When a social worker in Hong Kong consults a large language model about a child protection case, or a counsellor in Singapore uses an automated screening tool to triage clients, no clause in their professional code offers guidance on verifying outputs, disclosing tool use to clients, or assigning responsibility when an algorithm’s recommendation contributes to a harmful outcome. The codes simply do not contemplate these technologies as part of practice.

The second is what the authors call a cultural relational gap, and it is arguably the more profound problem. The codes analyzed embed a set of assumptions that trace their lineage to Western liberal moral philosophy: that the client is a self-determining individual whose autonomy should be respected above other considerations, that the professional should remain impartial and emotionally detached, and that clear, firm boundaries must separate the helper from the helped. These assumptions, the researchers argue, do not align well with the relational ethics through which practitioners in Chinese-influenced societies actually navigate their work. Concepts such as guanxi (關係), the web of reciprocal relationships and obligations; xinren (信任), trust built gradually through relational demonstration rather than formal contract; and renqing (人情), the exchange of favors and emotional indebtedness that structures social life, are central to how helping relationships are formed and sustained in Hong Kong and Singapore. A practitioner who insists on strict impartiality and rigid boundaries may fail to build the very trust that makes intervention effective, yet a practitioner who engages relationally risks transgressing the code’s boundary rules. The codes, in other words, may already be in tension with everyday practice before artificial intelligence even enters the picture.

The third is a paradigm gap. Existing codes assume that professional practice can be separated into distinct modalities: face-to-face sessions, telephone contact, written correspondence, each governed by its own rules. Contemporary practice, however, is a continuous movement across hybrid digital-physical configurations. A home visit might begin with a video call, continue over a messaging app, and conclude with an in-person meeting; a counselling relationship might live partly in a chat thread and partly in a consulting room. The COVID-19 pandemic accelerated this shift dramatically, as researchers studying child protection social work during that period documented how practitioners developed what has been termed digital intimacy—forms of therapeutic and protective connection mediated through screens. The codes’ assumption of separable modalities cannot capture this continuous hybridity, and the ethical questions it raises, such as how confidentiality operates across platforms or where a professional relationship legally and morally “exists,” fall through the cracks.

The study’s central and most striking contribution is the demonstration that these three gaps do not merely coexist; they compound one another. Prior research indicates that practitioners are already doing the labor of relational ethics in hybrid practice—negotiating trust, obligation, and care across digital and physical spaces—yet their codes offer no support whatsoever for this navigational work. The result is a double bind: professionals must improvise ethically in situations of unprecedented technological complexity while using ethical frameworks that neither address the technology nor reflect the cultural realities of their societies. When something goes wrong, the practitioner stands alone, formally accountable to a code that never anticipated the situation and informally bound by relational norms the code does not recognize.

The timing of the research is significant. Both Hong Kong and Singapore have made artificial intelligence a centerpiece of national and territorial policy. Singapore’s Smart Nation 2.0 initiative articulates a vision of a thriving digital future for all citizens, and the city-state has even published sector-specific guidance, such as a 2026 guide for using generative artificial intelligence in the legal sector. Hong Kong’s 2025 policy address likewise emphasizes deepening reforms and leveraging technological strengths. Yet, as the authors point out, while adjacent professions are receiving tailored guidance, the helping professions—whose work involves some of society’s most vulnerable people—remain without updated ethical infrastructure. Elsewhere, the picture is beginning to change: the British Association of Social Workers has issued a statement on social work and generative artificial intelligence, and scholars in the United States have proposed embedding generative artificial intelligence competencies into social work education accreditation standards. The contrast underscores how far the Hong Kong and Singapore codes lag behind the realities their practitioners face.

The research also situates itself within a broader and increasingly urgent debate about whose ethical frameworks govern artificial intelligence in Asian societies. Critics have described a form of cognitive imperialism in artificial intelligence, in which systems and the principles governing them encode Western epistemologies while marginalizing indigenous ones. Scholars comparing Chinese and European approaches to artificial intelligence ethics have documented genuinely different starting points, with Confucian traditions emphasizing role-based obligations and relational harmony rather than individual rights. Similar arguments have been made about the need for culturally grounded approaches to automated decision-making in India and about the experiences of disabled users in the Global South, for whom imported artificial intelligence systems can feel profoundly foreign. Against this backdrop, the question posed by the new study—whose ethics govern?—is not rhetorical. If professional codes in Hong Kong and Singapore are revised simply by bolting on generic artificial intelligence clauses derived from Western autonomy-based frameworks, the cultural relational gap will persist and may even widen, producing codes that are simultaneously more technologically current and less culturally legitimate.

The philosophical machinery behind the analysis draws on postphenomenological philosophy of technology, particularly the work of Don Ihde and Peter-Paul Verbeek, who argue that technologies are not neutral tools but actively shape the moral relationships between people. In this view, a video call is not merely a channel for a home visit; it mediates what the practitioner can see, hear, and feel, reshaping the caring relationship itself. The authors also draw on the concept of ethics work developed by Sarah Banks, which reframes everyday professional ethics not as the mechanical application of rules but as the ongoing, situated labor of moral deliberation. This framing explains why the paradigm gap matters so much: if ethics happens in the flow of hybrid practice rather than in discrete, rule-governed episodes, then a code that addresses only discrete episodes misses where ethics actually occurs. Approaches such as ethical-constructive technology assessment, which embeds ethical reflection in the design and adaptation of technologies rather than in after-the-fact checklists, offer one possible direction for reform.

The implications extend well beyond two cities. Social work and counselling are professions built on relationships, and artificial intelligence is now implicated in every stage of that relationship-building: intake algorithms decide who gets seen, predictive models flag which families warrant scrutiny, generative systems draft reports and simulate role-play for training practitioners, and conversational agents increasingly offer direct support to clients. If the ethical frameworks governing these technologies are culturally misaligned with the societies in which they are deployed, the harms will not be evenly distributed, and the trust on which helping professions depend may erode. The authors’ argument is ultimately a call for fundamental rethinking rather than incremental updating. The challenge that artificial intelligence poses to Asian professional practice, they conclude, is not merely a matter of adding new content to old codes. It is a question of whose ethical frameworks should govern professional life in Asian societies—and answering it requires centering the relational ethics of guanxi, xinren, and renqing, not treating them as deviations from a universal standard. For the practitioners of Hong Kong and Singapore, and for allied professions across Asia navigating the same terrain, the study makes clear that the status quo leaves them ethically unsupported at precisely the moment when support is most needed.