Background

Ethical concerns have long placed specific limitations on the study of the human brain. The ability to create human cerebral organoids (HCOs) as models for brain research has fostered hope among scientists of overcoming many of the obstacles that have previously hindered research on neurodevelopment and brain disorders [1,2,3]. Yet, the extent to which the broad and diverse field of HCOs, ranging from very rudimentary neural cell cultures to organoids modelling partial networks of the brain, deserves protection remains a matter for debate [4,5,6]. The current focus of the ethical discourse complicates the evaluation, as discussions are primarily concerned with future developments that organotypic cultures might undergo as the technology advances. This is especially relevant since stricter research regulations are required in case future HCOs develop capacities worth protecting [7, 8].

Current ethical considerations are thus intricately intertwined with epistemological questions about the potential capacities HCOs as models for brain research may develop as the technology advances. Above all, they revolve around the issue whether future HCOs will develop sentience or even consciousness, and thus have the ability to experience pain or pleasureFootnote 1 [7, 9,10,11,12]. This focus is problematic given the profound disagreement about the bodily prerequisites for conscious perceptions [13], and, not least, the measures suitable for evaluation [14]. In HCOs, evaluating perceptive capacities presents even greater challenges than in humans or animals because cell cultures ‘in a dish’ lack the ability to engage verbally or behaviourally in an assessment.

To bridge this gap, ethicists, legal scholars, and neuroscientists have turned to similarities (or resemblances) with the human brain in order to infer insights about HCOs [12, 15, 16]. These resemblances in structural or functional terms have mainly been established by scientists to justify the use of HCOs as models in brain research [17, 18]. Ethicists, however, draw on these similarities to the human brain to access the need for regulatory policies [4, 12, 14, 15, 19]. In ethical discourse, similarities at very early stages of cell development thus serve as a rationale for assumptions about the further development of HCOs as the technology advances. In this paper, I argue that such reasoning is unwarranted given the diversity of similarity relations that constitute HCOs as models for scientific practice. On this basis, I show that adopting a precautionary approach in this way risks disproportionately restricting research.

To substantiate my argument, Sect. “Human cerebral organoids as models in science” introduces the scientific practice of generating organotypic cultures with varying degrees of complexity and sophistication that resemble the human brain to the extent necessary for fulfilling specific research goals. Sect. "Similarity-based views in ethical discourse" engages with the ethical discourse, arguing that, given the heterogeneity of HCOs, it is not reasonable to assume that organotypic cultures will increasingly resemble the human brain as the technology advances. This perspective has far-reaching implications for advocates of the precautionary principle, particularly for those who propose relying on bodily surrogates of capacities deemed worthy of protection. In Sect. “Challenges in detecting early warning signs”, I address the challenges inherent in making reliable predictions about the future path organotypic cultures may take. Before proceeding, however, a more detailed understanding of the scientific efforts behind generating organotypic cultures is required in order to comprehend the epistemic value references to similarity might or might not have.

Human cerebral organoids as models in science

This section provides an overview of the scientific practice of creating organotypic cultures in the laboratory. The aim is to offer a sense of the procedure’s complexity, without in any way attempting to provide an in-depth analysis, which is beyond the scope of this paper.

In organoid research, scientists ‘build’ models, or rather model organisms, for experimental purposes.Footnote 2 To this end, they make use of pluripotent stem cells that have the potential to differentiate into a multitude of cell types. These cells by themselves, however, lack the inherent ability to guide differentiation towards any particular organ, rendering the artificial environment of the laboratory crucial for generating HCOs from cell cultures [23]. For this reason, researchers who generate HCOs as models for brain research invest considerable effort in crafting a suitable laboratory environment [24,25,26].

Despite the importance of the production process no uniform protocol exists for generating HCOs. Researchers make use of various ensembles of bioreactors, scaffolds, media and suspensions that constitute the laboratory environment. Broadly, two approaches can be distinguished based on the extent to which cell development is controlled during the production process. One is patterned protocols that closely guide cell development by substituting distinct growth factors at particular stages of the production process, yielding organoids that model specific brain regions like the forebrain in greater detail. A second approach relies on undirected protocols which make use of media and suspensions that merely provide the growth conditions for maturation. The latter kind of protocols controls developmental trajectories of cells to a lower extent and typically generates organoids that represent several brain regions in less detail [6, 27].

Regardless of the protocol used, the resulting organoids still need to be validated as research models. To establish their validity, scientists make use of comparisons to the human brain. In their own words, when generating an organoid for brain research the aim is to “closely mimic the endogenous developmental program” ([24], p. 2329) of human corticogenesis. Validation therefore involves establishing similarities to the human brain in terms of the variety of cell types, their organisational structure, and functional patterns [18, 24, 28,29,30]. One argument in favour of the validity of protocols thus is the observation that (“[…] cerebral cortical regions display an organization similar to the developing human brain at early stages”([26], p. 373). During maturation, organoids might even display borderline zones characteristic for particular stages of brain development that are “reminiscent of the mid–hindbrain boundary”([26], 374). Temporarily, the expression of such borderline zones may complicate the process of distinguishing single brain regions from each other. In functional terms, the “[…] gradual evolution of EP [electrophysiological] properties over development [is cited] that resembles hallmark features of the developing neonatal brain.” ([30], p. 859) Despite these similarities, HCOs also lack certain cell types, particularly vascular cells, which limits organoid complexity and a more advanced degree of maturation. This is because the absence of sufficient vascularization leads to necrotic cell death when the organoid reaches a certain size, affecting the ability of cell cultures ‘in a dish’ to form increasingly differentiated and complex circuits [31].

Nonetheless, the variety of protocols, which make use of heterogeneous laboratory constituents, provides researchers with the opportunity to establish similarities to the human brain in various ways and degrees. Thus, when generating an organoid, researchers select bioreactors and suspensions based on the brain region that the organoid should resemble in detail [31]. In the case of the forebrain, for instance, this means using patterned protocols that allow the substitution of distinct growth factors such as FOXG1 + or PAX6 + at particular developmental stages, as these stimulate cell differentiation in the desired direction [23]. As several research projects show [18, 26, 30], the degree of similarity to brain morphology or function displayed by neural cell cultures depends on the purpose for which HCOs are generated. This is illustrated by Qian and colleagues [18], who pursue the goal of generating an HCO for the study of Zika virus exposure. When applying a protocol in the production process the researchers prioritize resemblances in respect to the cortical differentiation of the forebrain, which is the primary region affected by the Zika virus in humans. Since the degree of sophistication when modelling the human brain depends on the selection of bioreactors, media, and growth factors, higher sophistication in one respect can only be achieved at the expense of being less precise in other respects [23].

Current research practices thus align well with Giere’s influential similarity account of models in scientific practice, particularly in terms of illustrating how researchers choose scientific models that resemble their research target in “relevant respects and degrees” ([32], p. 81). That material models represent an object of research by similarity constitutes a prevalent though not uncontested view in the philosophy of science [33,34,35].Footnote 3 In this regard, the observation that resemblances depend on scientists’ judgment explains that reference to similarity might have different epistemic value. Such an approach, however, introduces considerable variability among HCOs, which pose a challenge for an ethical discourse that relies on resemblance to evaluate moral issues that may arise.

Similarity-based views in ethical discourse

The current ethical discourse on the moral status attributable to HCOs is mainly concerned with the issues that may arise as the technology advances. Similarities to the human brain thereby serve as a means to evaluate the need for additional research regulations. Such resemblances therefore justify assumptions about the potential of HCOs to develop specific capacities, in particular, the ability for conscious perception [4, 12, 14].

In this context, some participants in the ethical discourse refer to “the genetic and developmental similarity to human brains” ([12], p. 9), which has largely been established by scientific practice, to justify HCOs’ use as models for brain research. However, these comparisons also shape regulatory approaches in the ethical discourse [4, 12, 15]. Ethical reasoning appears to presume that as technology advances, HCOs will increasingly resemble the human brain, thereby raising more and more complex moral issues. Of special importance is the study by Trujillo and colleagues because it reports similarities in the formation of functional circuits between HCOs and preterm infants that go beyond birth [36], a point in time at which conscious processing is generally assumed in humans [37].

The use of similarities to the human brain as indicators for conscious processing complicates matters due to the lack of a unified definition of consciousness and disagreements about the necessary bodily prerequisites for conscious perception [12, 13]. The main participants in this debate acknowledge that current HCOs are far from achieving human brain size or structure [6, 7, 11, 12]. However, there is consensus that ethical considerations should anticipate potential future developments that may arise [6, 7, 12, 19].

Given the significant uncertainty regarding the future trajectory of this biotechnology, precautionary reasoning is prevalent in current ethical debates [7, 12, 19, 38, 39]. Generally, the precautionary principle advocates for action in the face of potential threats even when scientific agreement on the issue is lacking [40]. Slight variations of this principle also propose a knowledge threshold or trigger condition that could raise the bar for evidential requirements before taking action [41]. However, once a decision to act is made, the selected measures remain open for reassessment, considering both short-term and long-term costs and benefits [40, 42].

Policy documents, treaties, and communications featuring precautionary reasoning can be traced back to the 1970s or even 1950s, when such reasoning was shaping debates on environmental threats and human health issues [40, 41]. The decision to act with ‘foresight’ in assessing potential (health) risks in the face of scientific disagreement expressed a normative choice based on the perceived level of certainty and severity of empirical risk, thereby prioritizing environmental protection and public welfare [41, 42].

In the precautionary discourse on organoids, normative considerations take centre stage. This is because the prospect of sentience or consciousness, crucial to moral consideration, shapes the debate on regulatory policies. While no agreement has yet been reached on any acceptable level of risk or evidence-threshold for caution [7, 12, 38, 39], several authors advocate evaluating precautionary measures under the assumption that HCOs might be capable of consciousness [7, 12]. Some of the participants in the discussion put the phenomenal experiences HCOs might have at the centre of the discussion while others consider even a minimal form of moral status to be justified only if neuroscientifically some form of consciousness seems likely [38]. In this context, Żuradzki cautions against an “over-attribution of moral status” ([43], p. 55) that fails to adequately consider the likely societal benefits of laboratory research.

Against this backdrop, Birch and Browning, as advocates of the precautionary principle, propose defining so called “neurological ‘warning signs’ of sentience” ([19], p. 57). The proposal implies looking for bodily surrogates of consciousness that may warrant caution. More precisely, they suggest searching for the neural correlates of conscious experience (NCCs) that have been associated with conscious processing in humans. The authors have suggested that: “if an organoid contains structures or mechanisms that any serious and credible theory of the human NCCs posits to be sufficient for conscious experience, we should take proportionate measures to regulate research on that organoid.” (ibd., italics in the original).

The aim of this suggestion is to set “the evidential bar for taking precautions at an intentionally low level.” (ibd.) To this end, a variety of brain regions (such as the forebrain), local and global mechanisms, as well as bi-directional connections (e.g. between thalamus and cortex) have been identified as potential warning signs. Yet, it can be doubted that the suggestion indeed sets a low threshold for the burden of proof for taking precautions. This is because even in humans, inferences about conscious states based solely on neural patterns remain a subject of debate [44, 45]. One of the main reasons is the correlational nature of data, which leaves open the question whether observable patterns indeed play a causal role in the emergence of consciousness or if they should be regarded as mere epiphenomena [46].

Birch and Browning’s suggestion, however, goes even further, proposing not merely resemblances but sufficient, if not exact, correspondence to the human brain to define legitimate warning signs, as brain regions or patterns that have, in humans, been associated with consciousness serve as surrogates. Given that HCOs generated through scientific practices resemble the human brain in various ways and degrees, setting such criteria for early warning signs may not be feasible. The challenges this poses for the evaluation of research regulations will be discussed in the final section.

Challenges in detecting early warning signs

Defining warning signs to evaluate research policy is a common practice among proponents of the precautionary principle. However, using neural patterns or regions associated with conscious processing in humans as surrogate markers complicates matters. In what follows, I outline three objections to the approach discussed in the previous section.

Ethical discourse surrounding HCOs aims to anticipate future moral issues that will arise as the technology advances [6, 12, 47]. However, moral considerations need to consider the existing, rudimentary states of HCOs used in research, as these pose a distinct challenge to the idea of defining early warning signs based on neural patterns. This is specifically the case because any attempt to assessing risks based on such rudimentary formations of brain patterns, above all, presupposes that these cell cultures follow a uniform course of development consistent with human corticogenesis.

Current research practices and, more particularly, the lack of a commonly agreed-upon protocol for HCO generation make it highly unlikely that organotypic cell cultures will follow a homogenous developmental path as the technology evolves. Given the diverse laboratory conditions and materials guiding HCO maturation, it is more plausible to assume that the developmental trajectories of these cultures will significantly vary. As a result, it becomes difficult to predict the future development of these cell cultures even if initial stages show concerning neural patterns.

Furthermore, and relatedly, using neural patterns as bodily surrogate markers for caution requires their reliable identification. The varied degrees of sophistication, differentiation, and interconnectivity among HCOs, however, complicate the identification of any neural pattern or structure previously associated with conscious processing. In the philosophical literature this issue has been addressed in terms of the “Boundary Problem” ([48], p. 378), i.e., the problem of reliably distinguishing neural mechanisms associated with targeted cognitive abilities from background processes. In the case of HCOs, this problem is exacerbated because during early development some regions in question lack distinct boundaries, making it difficult to identify relevant patterns [18, 26].

In this regard, the influence of diverse artificial laboratory environments on HCO development is particularly problematic. Nonetheless, the observation that changes in the artificial environment have a significant impact on the inner formation of HCOs corresponds well with current views on the very close relationship of a maturing organism with its environment. According to this view the surroundings do more to a maturing organism than merely maintaining and preserving autonomous growth. Instead, factors like metabolic exchange for homeostasis, temperature regulation, and functional circuits that allow for mutual feedback make the maturing organism an intrinsic part of its environment [49]. Such entanglements not only impact the early stages of organismal development but continue to influence their core structure as organisms mature [50].

However, even assuming that the neural patterns of interest are identifiable in HCOs, the question remains how to link the biological processes to the mental faculties which according to precautionary reasoning warrant caution.Footnote 4 This issue seems crucial since HCOs as artificial biological systems might display a “psychological architecture” ([13], p. 611) which significantly deviates from the ones in humans.Footnote 5 The possibility that consciousness arises in HCOs in a different manner than in humans, however, poses a significant challenge for an assessment.Footnote 6 Most importantly, however, the possibility that the emergence of consciousness in HCOs may be associated with different biological underpinnings than in humans carries the risk of incorrectly associating neural patterns with conscious processing.

In the philosophy of science, this risk of false association based on resemblance has been discussed in terms of mistargeting or misrepresentation. The criticism is part of broader arguments against similarity-based reasoning in science [34, 53]. In this particular case, however, incorrectly associating neural patterns with sentience or consciousness poses a significant problem. The reason is that the chance of mistargeting or misrepresentation could mean that HCOs displaying neural patterns of interest might in fact lack the potential for consciousness. The reverse is equally true, because even in the absence of the neural formations in question, HCOs might develop capacities ethical discourse deems worthy of protection.

Conclusion

This paper has critically examined prevailing assumptions in the precautionary discourse on brain organoids. The first is the notion that resemblances to human neural structures legitimate assumptions about the future trajectories HCOs might take. The second is the idea that similarities to human brain patterns, particularly to those associated with consciousness or cognitive abilities, should warrant caution.

Instead, I have shown that the heterogeneity with which scientists generate model organisms undermines assumptions about the epistemic value that resemblance might hold for ethical considerations. Specifically, scientists construct these models to exhibit similarities to the extent needed to achieve their research objectives. As such, initial resemblances may not necessarily indicate that organotypic cultures will evolve to a level of sophistication and complexity that would warrant ethical concern regarding potential mental capacities.

Raising awareness for the characteristics that constitute models in scientific practice may encourage a more pragmatic approach to dealing with the uncertainties surrounding HCOs as an emerging biotechnology. A critical aspect of this approach could involve reevaluating the time frame for assessing risks and potential harms. By narrowing the focus from long-term risks to more immediate societal benefits, we could reorient the discourse. Such a shift could not only illuminate the ethical considerations related to HCOs but also inform future deliberations on regulations needed for emerging biotechnologies. In this way, this analysis of the epistemic basis of the ethical discourse concerning HCOs may even enrich future negotiations on the regulatory frameworks essential for advancing biotechnologies responsibly.