The starting point for the analysis is a tension that runs through the entire field. On one side sits a companion animal framework, in which relationships between humans and social animals like dogs and horses are assumed to be potentially mutual: the animals enrich human lives while also benefiting from care, contact, and even the activity itself. From this perspective, AAS can be seen as promoting a good life for the animals involved. On the other side is a harm-minimization perspective that resembles how society views the use of animals in biomedical laboratory research. Here the animals are in human service, potentially at their own expense, and the ethical goal is to avoid pain, stress, and other harms while reducing and replacing animal use as far as possible. Under this framing, animal-assisted services at best represent an activity where the animals do not suffer but from which they do not meaningfully benefit.

These two framings map onto a conceptual distinction that is central to modern animal welfare science: the difference between negative and positive welfare. Negative welfare follows the principle of non-maleficence, the do-no-harm principle, and is typically assessed through indicators of acute or chronic stress, such as cortisol levels, heart rate variability, and behavioral signs of anxiety. Positive welfare, by contrast, requires evidence that animals are flourishing through predominantly positive mental states, the development of competence, and resilience. The authors point out that claims of positive welfare demand robust evidence of positive affective states, whereas the mere absence of harm or stress is sufficient only within a negative welfare framing. Confusing the two, they argue, is one of the most consequential errors they found in the literature.

The review’s methodology was deliberately interpretive rather than statistical. The team searched Web of Science and Scopus using terms built around animal-assisted interventions, service and therapy animals, and animal welfare, combined with dog or horse. After screening, 176 full-text articles were assessed and 116 met the inclusion criteria. The papers were sorted into three categories using a scoring system: direct welfare assessment based on animal-based behavioral, biological, and physiological indicators; indirect assessment through surveys, handler reports, and resource- or management-based indicators; and conceptual or ethical discussions including reviews. The sample split almost evenly between horse-focused papers (52) and dog-focused papers (56), with eight covering both species. Sixty papers measured welfare directly, 25 used indirect measures, and 31 were reviews or conceptually driven papers.

One of the review’s three key findings is that ethical frameworks in this research area are highly diverse and at times insufficiently specified. The authors catalogued an impressive range of reference points. The Five Domains model, which in contemporary versions recognizes the full spectrum of welfare states from negative to positive, appeared repeatedly. Quality of Life, originally developed in human health care and later adapted in veterinary medicine, served as an ethical anchor for several studies. The One Health approach, emphasizing the interconnections among human, animal, and planetary health, was invoked with varying degrees of specificity. A more demanding interpretation, drawing on the related One Welfare concept, holds that well-structured AAS can promote human and animal flourishing simultaneously, and that robust benefits for the animal are a minimum requirement for ethical justification rather than an optional bonus.

Perhaps the most striking borrowings come from laboratory animal ethics. A small number of studies explicitly applied the 3Rs framework, originally developed for animal research: Refinement to minimize harm, Reduction to limit animal involvement, and Replacement to consider alternatives. One study even borrowed the biological concept of symbiosis as a cost-benefit model for evaluating human-horse relations in equine-assisted services. Other papers introduced concepts like animal consent, animal agency, and self-efficacy, with one team contrasting consensual interaction with being part of an exploitative relationship. Legal instruments such as the German Animal Welfare Act and professional guidelines like the IAHAIO White Paper also served as ethical foundations in parts of the literature.

The second key finding concerns how these frameworks are applied, and where they break down. The authors identified cases of conceptual conflation when positive and negative welfare perspectives are combined without acknowledging their fundamentally different understandings of welfare. In several studies, positive welfare or related notions like animal wellbeing were introduced as guiding principles in the introduction, and conclusions were framed in those terms, yet the actual findings were limited to the absence or minimal occurrence of stress indicators or stable physiological measures. No substantial assessment of robust benefits to the animals was presented. Claims of mutual benefit thus coexisted with welfare assessments restricted to negative indicators, blurring the line between the absence of harm and the presence of positive welfare. The authors suggest a practical explanation: measures of negative welfare are more established and easier to operationalize, even as the field has shifted rhetorically toward positive welfare and mutuality.

The rationales researchers gave for studying animal welfare revealed a further layer of framing. Only 13 papers, or 11.2 percent, articulated rationales centered on the animals’ own interests, while 8 papers, or 6.9 percent, focused predominantly on human concerns such as improving animal performance, therapeutic outcomes, client safety, and organizational efficiency. Another 32 papers, 27.6 percent, combined both perspectives. Strikingly, in 63 papers, 54.3 percent, the underlying rationale could not be clearly identified. The authors are careful to note that this does not indicate an absence of ethical reasoning, but rather that the rationale was not explicit enough to classify, which itself undermines transparency and comparability across the field.

The third key finding is a recurring relational framing across the literature. A triadic handler-animal-client model appeared repeatedly as a framework for discussing responsibility, risk, and cooperation: who recognizes and responds to the animal’s needs, how client-animal interactions are managed, and how the quality of each relationship within the triad affects welfare. Studies emphasized the handler’s duty to advocate for the animal, including ensuring the animal can withdraw from interactions. The authors argue that this relational turn, while valuable, requires scrutiny. They note that the broader relational perspective in human-animal studies has been described as a mixed blessing: it can decenter human interests and emphasize interdependence, but it may also obscure asymmetries of power. As one cited review of therapy dog welfare puts it, handlers must recognize the power differential within the triadic relationship and protect the dog’s welfare despite the preferences of recipients or handlers.

The review’s language analysis adds a vivid dimension to these findings. Word choices in the literature encoded normative assumptions about the animals’ roles. Some studies described animals as a therapeutic tool, while others framed them as co-therapists, social butterflies, or emotional mediators. The word use was employed descriptively by some authors and pointedly placed in quotation marks by others. Therapy-first language, which foregrounds the licensed therapist rather than the animal, was recommended by some researchers as a corrective. The authors conclude with three recommendations: clearer articulation of ethical framings so studies can be positioned along key ethical axes, consistent application of framings throughout a paper so that absence of negative findings is not mistaken for evidence of positive welfare, and the inclusion of reflexivity statements in which authors disclose the normative assumptions they bring to their research. Greater transparency, they argue, will improve the interpretation, comparability, and communication of welfare findings, and provide a more robust foundation for guidelines and policies aimed at improving, and potentially rethinking, animal-assisted services from animal-centered perspectives.

Subject of Research: Ethical framings in animal welfare research on dogs and horses in animal-assisted services

Article Title: Partners or tools? Ethical framings shape welfare research on dogs and horses in animal-assisted services

Article References: Zenker, F., Thodberg, K., & Sandøe, P. (2026). Partners or tools? Ethical framings shape welfare research on dogs and horses in animal-assisted services. Frontiers in Veterinary Science, 13, Article 1960144. https://doi.org/10.3389/fvets.2026.1960144

Image Credits: AI Generated

DOI: 10.3389/fvets.2026.1960144

Keywords: animal-assisted services, animal welfare, animal ethics, therapy dogs, equine-assisted therapy, positive welfare, One Health, Five Domains model, 3Rs framework, relational ethics, critical review, veterinary science