Ethical implications of visual neuroprostheses—a systematic review

Objective. The aim of this review was to systematically identify the ethical implications of visual neuroprostheses. Approach. A systematic search was performed in both PubMed and Embase using a search string that combined synonyms for visual neuroprostheses, brain–computer interfaces (BCIs), cochlear implants (CIs), and ethics. We chose to include literature on BCIs and CIs, because of their ethically relavant similarities and functional parallels with visual neuroprostheses. Main results. We included 84 articles in total. Six focused specifically on visual prostheses. The other articles focused more broadly on neurotechnologies, on BCIs or CIs. We identified 169 ethical implications that have been categorized under seven main themes: (a) benefits for health and well-being; (b) harm and risk; (c) autonomy; (d) societal effects; (e) clinical research; (f) regulation and governance; and (g) involvement of experts, patients and the public. Significance. The development and clinical use of visual neuroprostheses is accompanied by ethical issues that should be considered early in the technological development process. Though there is ample literature on the ethical implications of other types of neuroprostheses, such as motor neuroprostheses and CIs, there is a significant gap in the literature regarding the ethical implications of visual neuroprostheses. Our findings can serve as a starting point for further research and normative analysis.


Introduction
Visual neuroprostheses are sensory prostheses that can be used to restore impaired vision (see figure 1). The need for such a prosthesis is illustrated by the high global burden of disease due to impaired vision. In 2015, an estimated 36 million people were blind throughout the world, and approximately 216.6 million people had a moderate to severe visual impairment (Bourne et al 2017). Many causes of blindness and visual impairment, including degenerative disease, diabetic retinopathy, glaucoma or trauma, are known to have hardly any treatments.
Visual neuroprostheses have the potential to partially restore vision in these individuals (Shepherd et al 2013). These devices are part of the spectrum of neuroprostheses that interact with the nervous system to restore sensory or motor function (Schwartz 2004, Adewole et al 2016, Glannon 2016, Wright et al 2016. Visual prostheses aim to restore vision by targeting one of several visual processing areas, including the retina, the optic nerve, the lateral geniculate nucleus, and the visual cortex (Shepherd et al 2013, Lewis et al 2015, Mirochnik and Pezaris 2019, Niketeghad and Pouratian 2019. Research on all of these approaches is on-going, with the emergence of several clinical trials and significant growth in the field of neuroprostheses in general over the past decade (Mirochnik and Pezaris 2019).
The development and use of visual neuroprostheses are paired with ethical implications. Some ethical implications are unique for visual neuroprostheses, while other ethical implications are applicable to neuroprostheses in general, such as the informed consent process. Some ethical implications of visual neuroprostheses are similar to those of other sensory neuroprostheses, such as some brain-computer interfaces (BCIs) and cochlear implants (CIs). CIs are a more mature technology and interact with the peripheral nervous system, and are hence accompanied by lower risks and better restoration of function. This way, they could provide a bench mark for other forms of neurotechnology.
To date, a comprehensive overview of the ethical implications of visual neuroprostheses is missing. The aim of this study is to systematically identify the ethical implications of visual neuroprostheses from the academic literature.

Design
In order to identify the ethical implications of visual neuroprostheses, we conducted a systematic review of the academic literature. The review process was performed following the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines (Moher et al 2009). Ethical implications were broadly understood to be expressions where ethical reasons, principles and values are at stake, or any type of ethical or societal issue or consideration, which may not necessarily take the form of a fully argument-based reason (Mertz et al 2016).

Search strategy
The literature search of the Pubmed and Embase databases was conducted in October 2019, and we conducted an additional search in May 2021 to include articles published since the original search. These databases were selected for their comprehensiveness and the review's focus on medical-ethical implications. The choice of databases was discussed with an experienced librarian from the Utrecht University Library. The search strategy combined synonyms for visual prostheses, CIs, neuroprostheses, neurotechnologies, BCIs, and ethics (see supplementary files 1-4 available online at stacks.iop.org/JNE/19/026055/mmedia).

Study selection and inclusion criteria
Articles that focused on the ethics of visual neuroprostheses, CIs, BCIs, and neurotechnologies in general were included (see figure 2). We included literature on BCIs and CIs because of ethically relevant similarities and functional parallels to visual neuroprostheses. We included literature on neurotechnology in general because of ethically relevant parallels to the field of visual neuroprostheses. We excluded articles that were not written in English or Dutch, that were published before 2009 and or lacked an abstract. In addition, articles that described very specific (sensori-motor) neurotechnologies or BCIs that were not relevant for visual neuroprostheses were excluded based on their title and abstract. For example, the ethical implications of psychiatric therapeutic applications of BCIs were considered to diverge too much from those of sensory therapeutic applications. Title and abstract-screening (tiab screening) was performed by two researchers (EAMV and KRJ).
Following tiab screening, EAMV & KRJ conducted a full-text screening of all the articles that were included (see figure 2). The results of the fulltext screening were checked by a second researcher (EAMV, DRH or KRJ). We excluded articles from the full-text analysis if the full text was not available, if they discussed specific (e.g. motor-function) neurotechnologies, or BCIs that were not relevant for visual neuroprostheses, or if they did not mention any ethical issues. Only articles from peerreviewed journals were included. In both tiab screening and full-text screening the results were juxtaposed to ensure consistency. In case of discrepancies between researchers, final inclusion was based on consensus.

Data extraction and analysis
A data extraction sheet was used to extract data and identify ethical implications systematically (see supplemental file 5). We extracted general information about the articles, including the author names, year of publication, country of origin, the authors' affiliations, the device or technology mentioned, and the aim of the article. The data extraction of all the articles was performed by EAMV or KJ and checked by a second reviewer (EAMV, DRH or KRJ). Discrepancies in the extracted data were discussed and resolved within the research team by consensus.
Ethical implications from all articles were then analyzed by assigning a subcode (subtheme) to each implication. Next, a list of subcodes was generated. When articles addressed the same topic, these implications were categorized under the same subcode. For each ethical implication we noted which article(s) included that topic and how many articles mentioned the particular topic. Related subcodes were categorized into main codes, in order to provide an overview of broader ethical implications. The ethical implications, subcodes and main codes were based on consensus within our research team. The formulation of these ethical implications, subcodes and main codes was re-evaluated among all the researchers several times, to ensure that all the extracted data was categorized by a code that best described the ethical implication.

Search and selection
The database searches resulted in a total of 1521 records. After the removal of duplicates, 1133 records remained. After tiab screening and full-text screening, 84 articles were included for data extraction and analysis (see figure 2).

Characteristics of included articles
In total 84 papers were included, of which six focused specifically on visual neuroprostheses, 24 on neurotechnology in general, six on neuroprostheses in general, 27 on BCIs specifically, and 21 on CIs specifically (see table 1).

Ethical implications
We identified 169 ethical implications in total, which have been categorized into seven main themes: (a) benefits for health and wellbeing; (b) harm and risk; (c) autonomy; (d) societal effects; (e) clinical research; (f) regulation and governance; and (g) involvement of experts, patients and the public.

Benefits for health and well-being
Twenty-four implications related to the theme 'Benefits for health and well-being' . An overview of these implications can be found in supplemental file 6.
First, restoration of function is often mentioned as an important health benefit. Repair and restoration of normal function is considered to be the primary goal of neuroprostheses (Tbalvandany et al 2019). However, while stimulation via neuroprostheses can provide some restoration of sensory function, complete restoration is not (yet) possible (Lane et al 2011, Laryionava and Gross 2011, Lucas 2012, Xia and Ren 2013, Rosenfeld and Wong 2017, Ereifej et al 2019, Wood et al 2019, Coin and Dubljević 2020. Moreover, several authors argued that what would be considered 'meaningful restoration' varies among people and also differs between persons who are sighted and those who are blind or visually impaired (Lane et al 2011, Tbalvandany et al 2019, Hansson 2020. Restoration of function is arguably not important to all of those who might be eligible for neuroprostheses and therefore not all of them will be interested in such restorative devices (Glannon 2016, Burwell et al 2017. Moreover, the literature on CIs showed that the degree of restoration of function differs between users (Kermit 2012, Pass andGraber 2015). Furthermore, measurement of the functional improvement of neuroprostheses is difficult, and should be improved by including qualitative aspects and actively involving (potential) users of the technology (Lane et al 2012).
Second, rehabilitation was mentioned several times in the literature. Rehabilitation and training to use a device were argued to be crucial, yet difficult or strenuous for most recipients and some might never learn to use the device (Lane et al 2011, Mikołajewski and Mikołajewski 2013, Xia and Ren 2013, Carmichael and Carmichael 2014 To describe the different types of brain repair strategies being developed in basic and clinical research, and provide a brief overview of recent advances in AI that have the potential to improve neuroprostheses. Pass and Graber (2015) USA Perspective

Cochlear implants
To argue that the informed consent process for cochlear implantation must include access to deaf perspectives.
Postan ( To understand how BCIs may harm or help the user, and to investigate whether ethical guidance is needed. Schermer (2009) The Netherlands

Perspective BCIs
To focus on the effects that new neurotechnologies may have on our 'symbolic order' and on the ways in which popular categories and concepts may change or be reinterpreted because of emerging neurotechnologies.

Neurotechnologies
To review different stakeholders' opinions and expectations, compile information on the state-of-the art in medical neurotechnology and describe and assess the technologies that are used to build these neurological implants.
(Continued.) Vlek et al (2012) The Netherlands Other BCIs To report the results of a panel that identified ethical issues related to BCIs in four case scenarios during the 4th International BCI meeting and illustrate where treatment and research interests conflict, and ethical concerns arise.
Weisleder (2012) USA Other Visual neuroprostheses To present interviews with three representatives for the blind regarding 'blind culture' .
Wolbring ( comes with the challenge of adjusting to novel senses and experiences (Glannon 2016) and despite extensive user training, BCIs can be frustratingly slow (Sample et al 2019). Others mentioned that the incorporation of a device requires regained functionality and trust in the tool (Tbalvandany et al 2019). Furthermore, intensive training may impose time, physical, emotional, and financial burdens on the user and their family (Byrd et al 2011, Pass and Graber 2015, Burwell et al 2017. Hence, further clinical research on training and rehabilitation of patients with neuroprostheses is needed (Mikołajewski and Mikołajewski 2013). Third, specific ethical implications for restoration and rehabilitation were raised regarding children. The CI literature, for example, suggested that early implantation of CIs in children may increase the chance of restoration, hence the option of postponing surgery until a child can make its own decisions requires a careful weighing of benefits and risks (Hardonk et al 2010, McCormick 2010, Byrd et al 2011, Melton and Backous 2011, Lucas 2012, Teagle 2012, Lee 2016, Owoc et al 2018. Furthermore, rehabilitation challenges faced by children with CIs are more severe than those faced by adults. Where rehabilitation challenges of CIs in adults are primarily to learn to hear again, children are faced with additional challenges, such as having to learn how to demonstrate various competences (e.g. social and linguistic) as they develop (Kermit 2012). Moreover, the rehabilitation process following CI implantation may impact the child's self-perception (Kermit 2012). Support from family members and parents is considered to be crucial for successful (re)habilitation of children (Kumar et al 2017).
Fourth, various authors mentioned benefits of neuroprostheses in well-being. For example, authors argued that the ultimate goal of neuroprostheses is to give the patient the feeling of 'belonging somewhere again, as being part of life' (Slatman as cited by Tbal- . BCIs were mentioned to contribute to wellbeing by enhancing communication (Sample et al 2019). Yet, in the case of CI implantation in children, well-being could also be negatively affected. Concerns were expressed that when normal hearing levels are not fully attained, children may suffer comparing themselves to children with normal hearing (McCormick 2010). Some users of CIs may request device removal, due to dissatisfaction with the technology (Owoc et al 2018). For visual prostheses, one paper mentioned that psychological, social, and cultural factors associated with vision have to be considered together with the ability to see, in order to assess the overall value of the prosthesis (Glannon 2016).
Lastly, CIs are argued to improve one's sense of safety-for example, by allowing CI users to pick up auditory cues from their surroundings (Byrd et al 2011, Maki-Torkko et al 2015, Vieira et al 2018.

Harm and risk
Aside from benefits for health and well-being, the use of visual neuroprostheses comes with potential risks. An overview of 34 implications identified in the literature can be found supplemental file 7.
First, side-effects and unintended effects were mentioned. Implantation of a neuroprosthesis comes with mechanical harms and bio-compatibility risks, including foreign body reactions such as inflammation, encapsulation, and glial scarring (Demetriades et al 2010, Kotchetkov et  Furthermore, the heat produced by the device is a potential risk for thermal damage to the brain (Klein and Ojemann 2016, Mitrasinovic et al 2018, Stieglitz 2019. Aside from mechanical harm and bio-compatibility issues, there may also be other side-effects and unintended effects of implantation. Neuroprostheses may have an impact on brain function or on other health problems (Lane et al 2012, Chari et al 2021. Furthermore, users may experience psychological difficulties due to changes in their life after implantation with a sensory neuroprosthesis (Lane et al 2011, Xia and Ren 2013, Glannon 2016, Klein 2016, Francis-Auton et al 2020, Hansson 2020. Furthermore, some authors are worried that users of neuroprostheses may experience changes in personality (Schermer 2009, Laryionava and Gross 2011, Lane et al 2012, Jebari and Hansson 2013, Mikołajewski and Mikołajewski 2013, Goering and Yuste 2016, Moritz et al 2016, Mudgal et al 2020. However, one author argued that such statements about personality changes due to neurotechnology may be exaggerated and should not be taken at face value (Hansson 2020). Others were concerned that neuroprostheses could affect one's sense of self, thereby raising questions about authenticity may potentially be abused, in the sense that they may be used for harmful purposes or become difficult to control (Schermer 2009, Demetriades et al 2010, Kotchetkov et al 2010, Mikołajewski and Mikołajewski 2013, Xia and Ren 2013, Keskinbora and Keskinbora 2018, Hendriks et al 2019, DeFranco et al 2020, Hansson 2020. Similarly, the possibility of dual use, the application of neuroprostheses for non-therapeutic purposes, such as the use of neurotechnology for military combat applications, may be hard to control and even be harmful (Demetriades et al 2010, Kotchetkov et al 2010, Tracey and Flower 2014, Farah 2015, Goering and Yuste 2016, Ienca et al 2018, DeFranco et al 2020.
Second, the stability and quality of the device may pose an ethical challenge. The devices that are currently available may lose functionality over time (Kotchetkov et  Third, device manufacturers may introduce biases into the devices that they develop, which may lead to potential harm and unequal outcomes for patients and research (Glannon 2016, Hendriks et al 2019, Hansson 2020.
Fourth, privacy issues form an important aspect of the harm and risks that surround neuroprostheses. Privacy issues depend on the type of neuroprostheses being used (Gladden 2017). Several authors mentioned that neuroprostheses require proper security with regard to information privacy, encryption, and data protection (Jebari andHansson 2013, Klein 2016 Fifth, a number of ethical implications related to epistemic uncertainty were mentioned. Various authors argued that there is quite some uncertainty and unknown risks associated with neuroprostheses: the long-term safety of neuroprostheses is poorly understood (Laryionava and Gross 2011, Vaughan and Wolpaw 2011, Jebari and Hansson 2013, Hendriks et al 2019, Reilly 2020) and for long term usability, further knowledge on requirements for reliable and stable devices is needed (Vaughan and Wolpaw 2011, Klein 2016, Moritz et al 2016, Stieglitz 2019). In addition, improvement and validation of surgical techniques for implants are required (Hansson 2020). Neuro-technological therapy and neuro-enhancement are still accompanied by many uncertainties and unknown risks (Demetriades et al 2010, Farah 2015, Klein 2016, Keskinbora and Keskinbora 2018, Chari et al 2021. For example, while wireless power transfer is emerging, much remains unknown in terms of (longterm) safety and long-term functionality (Klein 2016, Klein and Ojemann 2016, Rosenfeld and Wong 2017, Mitrasinovic et al 2018. One author argued that devices for therapeutic benefit should be designed in a way that mimics healthy neurodevelopment, which requires solid understanding of these mechanisms (Lucas 2012). Furthermore, the expected benefits of neuroprostheses have been suggested to lack solid evidence (Burwell et al 2017, Lazaro-Munoz et al 2018. Sixth, issues are mentioned that relate to high expectations and hype around neurotechnology. Patients, the public, and the media may have unrealistic and high expectations of neurotechnology, which could generate false hope and have negative effects such as disappointment (Haselager et al 2009, Schermer 2009 One author remarked that a risk-benefit analysis is a complicated endeavour, as benefits and risks are hard to quantify for hearing devices (Wolbring 2013).
Last, harm and risk concerns are especially precarious when neurotechnologies are applied to children. More research is needed to better understand the long-term stability and psychological effects of neuroprostheses, as well as their implantation in a developing nervous system (Mikołajewski andMikołajewski 2013, Pass andGraber 2015). According to the literature on CIs, CIs in children are safe and reliable (Melton and Backous 2011).

Autonomy
Eleven ethical aspects related to the theme 'autonomy' were mentioned. An overview of these implications can be found in supplemental file 8.
Autonomy is a multifaceted concept that was interpreted by various authors in different ways, using concepts such as liberty, independence and authenticity.

Societal effects
Visual prostheses can have far-reaching societal effects. An overview of 26 implications can be found in supplemental file 9.
First, neurotechnology can affect social participation. Several authors mentioned that restoration through neurotechnology contributes to social participation (Lane et  In the context of CIs, it was mentioned that parents may experience stress from having to choose whether their child should be implanted with a CI or not (Hardonk et al 2010).
Third, neuroprostheses influence our ideas about norms and normality. For instance, neurotechnology could alter the way in which we see ourselves, and raise questions about the relationship between humans and machines (Schermer 2009 . For visual prostheses specifically, it was mentioned that the focus on restoration of vision may promote the perception that the life of a blind person is worth less than that of a sighted person (Lane et al 2012) or that disability should be understood as a social construct that only occurs when one does not have access to restorative technology (Lane et al 2012). With regard to stigma, some authors argued that that neuroprostheses could increase social stigmas due to differences between the bodies of neuroprostheses users and non-users (Jebari 2013, Wolbring 2013, Aas and Wasserman 2016, Burwell et al 2017, Sample et al 2019 and that such stigma may influence the acceptance of neuroprostheses in society (Carter et al 2019). The use of neuroprostheses to enhance, rather than solely to restore 'normal' body functions, was frequently mentioned as an ethical implication (Schermer 2009, Demetriades et  A slippery slope argument stated that the acceptance of less controversial technologies may lead to the introduction of more controversial technologies in the future (Jebari 2013). Yet, one author argued that enhancement in the context of CIs might not be a real threat, as the devices are standardized and cannot be modified at will to perform beyond the restoration of 'normal' hearing (Lee 2016).
Fourth, several implications related to culture were mentioned. In the literature on CIs specifically, it was mentioned that CIs were considered a threat to Deaf culture, Deaf culture was argued to be inherently valuable ( Opposed to that, one author suggested that CIs are not necessarily culturally oppressive, as the decision to opt for a CI is equally culturally mediated to the argument to preserve Deaf culture (Lee 2016). With regard to visual prostheses, it was argued that the blind community may not have a 'blind culture' that is comparable to Deaf culture as described in the CI literature (Lane et al 2012, Weisleder 2012, Hansson 2020. Fifth, several cost implications were mentioned. Neurotechnologies may be disproportionally more expensive than traditional therapy, care, and rehabilitation (McCormick 2010, Miziara et al 2012, Mikołajewski and Mikołajewski 2013. The development of safe neuroprostheses is expensive and requires the realization of clinically and economically viable models to realize dissemination and implementation (Vaughan andWolpaw 2011, Stieglitz 2019). However, neurotechnologies may have a positive impact on the economy (Jebari and Hansson 2013). Furthermore, coverage by national health insurances will depend on the balance between benefit and cost (Thébaut 2013, Reilly 2020). Moreover, as there are inter-individual differences between BCI users, a well-functioning prosthesis requires the technologies to be tailored to individual users. However, this tailoring is costly and does not benefit other users (Vlek et al 2012).

Clinical research
An overview of 42 ethical implications in conducting clinical research on visual neuroprostheses can be found in supplemental file 10.
First, several key aspects of the informed consent process were mentioned in the literature. These implications referred to informed consent in both research and clinical settings, as was often seen in the literature on CIs. Research participants and patients need more than solely technical information when providing informed consent and differ in their information needs and preferences. Hence, sufficient time to ask questions should be provided (Lane et al 2011, Duarte et al 2015, Maki-Torkko et al 2015, Klein and Ojemann 2016, Rosenfeld and Wong 2017, Hendriks et al 2019, Francis-Auton et al 2020.
Furthermore, risks and possible benefits have to be explained carefully in the informed consent process of a research trial or clinical procedure (Hardonk et al 2011, Lane et al 2012, Hendriks et al 2019, Jeong et al 2019, Hansson 2020. Also, research participants consider ancillary and due care to be important aspects of the consent process (Lane et al 2011(Lane et al , 2012(Lane et al , 2016. Beliefs about normality and enhancement should be included in the informed consent process, because these may influence the perception of parties involved in care and policy-making, including medical insurance companies (Klein 2016). Due to the degree of complexity and amount of information relevant for the informed consent process, multiple information sessions are recommended to ensure that the research participant has understood the information (Lane et al 2012, Xia and Ren 2013, Klein and Ojemann 2016). Yet, epistemic uncertainties surrounding neuroprostheses can make it difficult to provide sufficient and understandable information within the informed consent process in both research trials and clinical procedures (Kermit 2012, Vlek et al 2012, Xia and Ren 2013, Klein 2016, Klein and Ojemann 2016. It was mentioned that parents generally rely on information provided by medical experts when deciding whether to implant their child with a CI (Hardonk et al 2010(Hardonk et al , 2011. Others favored the 'reasonable person standard' for the informed consent process (information that a reasonable person may need to make an informed decision should be included), and insisted that the informed consent process in the context of CIs requires access to the testimony of patient experiences (Pass and Graber 2015). Some authors argued that potential research participants may want to involve significant others or consult other patients on their decision to participate in a research trial (Lane et al 2011(Lane et al , 2012(Lane et al , 2016. Furthermore, it has been argued that consent of significant others should also be sought, as neurotechnologies also affect them (Klein and Ojemann 2016). The literature also described that it is important that the informed consent process prepares participants for the potential disruption of agency and identity (Klein 2016, Klein andOjemann 2016). Regarding factors that may influence the informed consent process, it was argued that pervasive attitudes about disability and stigma may be oppressive and thereby influence voluntary informed consent (Klein 2016). Also, unrealistic expectations about neuroprostheses can hamper informed decision-making and can lead to therapeutic misconception (Vlek et al 2012, Carmichael and Carmichael 2014, Klein 2016, Klein and Ojemann 2016, Burwell et al 2017. Furthermore, potential benefits and risks of participation in clinical trials and clinical procedures (including cultural and psychological aspects) should be carefully weighed, especially when the research participant or patient is a child (Byrd et al 2011, Lucas 2012, Miziara et al 2012, Chari et al 2021.
Second, several motivations of participants in research trials were described. Some referred to personal reasons. For example, participants may consider research participation to be consistent with their adventurous preferences or curious and/or risktaking nature (Lane et al 2011(Lane et al , 2012(Lane et al , 2016. Also, the desire for 'leaving a legacy' and being a pioneer could motivate research participation (Lane et al 2011(Lane et al , 2016. Increased access to healthcare or financial reimbursement were also mentioned as motivations for participating in clinical trials (Xia and Ren 2013). Authors mentioned that health benefits and independence can motivate participants (Lane et al 2011, Xia and Ren 2013. However, though the hope of receiving benefits from visual prostheses may be motivating, it may also be unrealistic (Xia andRen 2013, Hansson 2020). Participation of children in clinical CI trials could be motivated by the parents' hope that their deaf children would gain education or employment opportunities (Hardonk et al 2011). Furthermore, negative experiences, such as social stereotyping and prejudices of 'disability,' may motivate people to participate in clinical trials (Lane et al 2011). Guilt and desire for restoration may be poor motivations for participation as this indicates a poor adjustment to blindness (Lane et al 2016). Altruistic reasons, such as the wish to contribute to scientific progress, were also mentioned as motivation for research participation (Lane et al 2011, Xia and Ren 2013. Finally, it was argued that careful appeal to motivational factors is prudent, to prevent exploitation of potential participants (Lane et al 2012, Hendriks et al 2019, Hansson 2020. Third, several implications for recruitment, selection and screening were described. With regards to recruitment, it was suggested that experts from different fields (e.g. neuroscience, engineering, ophthalmology and rehabilitative medicine) should be included in the recruitment process, to ensure that prosthesis implantation is indeed the best option for each patient (Xia and Ren 2013, Hendriks et al 2019, Chari et al 2021. Both participant screening and recruitment should be guided by ethical guidelines (Xia and Ren 2013). The literature also suggested that it is difficult to recruit volunteers for research trials of visual neuroprostheses, as the benefits are still uncertain (Xia andRen 2013, Hansson 2020). Others, however, worried that candidates with severe disabilities are more likely to accept increased risks in BCI research trials, in the hopes of some minimal benefit (Burwell et al 2017). It was also noted that for those persons that still have some sight left, the risk of losing any remaining vision outweighs the potential benefits (Lane et al 2011(Lane et al , 2016. However, recruitment of participants with some residual vision can aid investigations of the effectiveness of the device (Lane et al 2016). The challenge is to balance respect for the participants' autonomy and the duty not to harm. This is particularly difficult in vulnerable groups who may be eager to participate (Lane et al 2011, Xia and Ren 2013, Klein and Ojemann 2016. With regards to the screening process of research participants, a psychologist should carry out an eligibility assessment to assess a candidate's suitability for the trial, including their motivation for participation, emotional state, potential impact of residual perception loss, and expectations regarding the efficacy of the device (Schermer 2009, Xia and Ren 2013, Lane et al 2016. Screening is critical for the success of clinical trials, especially in the early phases of such trials or when new devices are being tested (Xia and Ren 2013). Desire for withdrawal from the study always needs to be respected (Klein and Ojemann 2016). One article also stated that selection of participants should not be influenced by who is able to afford the medical costs of participating in such trials (Glannon 2016). Others mentioned that the presence of a contraindication is not necessarily a barrier for implantation of a visual prosthesis in a therapeutic setting, but that in these early experimental stages, (fully) eligible patients are preferred, due to uncertainties of risks (Xia and Ren 2013).
Fourth, when it comes to research with no prospect of therapeutic benefit, it is sometimes difficult to determine what constitutes an acceptable risk (Hendriks et al 2019).
Last, several authors mentioned post-trial access and post-trial healthcare costs. Furthermore, the technology may advance at a rapid pace, and due to the invasiveness of the procedure, participants in early proof-of-concept studies may not be able to participate in future studies or benefit from the use of next-generation devices (Klein 2016

Regulation and governance
Several papers stressed the need for regulation and legal and ethical guidance. An overview of the 16 implications of this main theme can be found in supplemental file 11.
First, several authors argued that legal guidance or regulation has to be in place in order to ensure the safety of neuroprostheses and to protect their users (Demetriades et al 2010, Jebari and Hansson 2013, Mikołajewski and Mikołajewski 2013, Tracey and Flower 2014, Glannon 2016  It was argued that neuroprostheses should be legally conceptualized as a part of the body (Schermer 2009, Sample et al 2019 and that the focus on neuroprostheses as an 'object' in regulation is too narrow and misunderstands the fact that the drive for innovation originates in the desire to improve health (Quigley and Ayihongbe 2018).
Second, ethical oversight was considered necessary for the governance of these technologies. Several articles mentioned that continuous ethical guidance and oversight is required to ensure that research is improved and to ensure that the technology fulfills societal goals (Jebari and Hansson 2013, Mikołajewski and Mikołajewski 2013, Carmichael and Carmichael 2014, Goering and Yuste 2016, Rosenfeld and Wong 2017, Carter et al 2019, Stieglitz 2019, DeFranco et al 2020. Ethical oversight of the application of neuroprostheses in children introduces specific additional questions, as traditional bioethical frameworks are argued to be insufficient (Lucas 2012). Moreover, further study is needed to guard against the potential downsides of rapid technological development (Mikołajewski andMikołajewski 2013, Kellmeyer 2018). Individual differences between users make it challenging to develop a standardized framework for protection and oversight (Lane et al 2016). Furthermore, as the community working on visual restoration is small, oversight and monitoring boards may have a conflict of interest (Lane et al 2016). Potential conflicts of interest may also arise with the development and clinical research of neuroprostheses, due to financial interests of researchers or the influence of commercial organizations or big companies (Xia and Ren 2013, Glannon 2016, Kellmeyer 2019, Hansson 2020, Chari et al 2021. Additionally, some authors hinted that ethical guidance offers limited benefits, as it may not always lead to a better participant experience (Lane et al 2016).
Last, it was mentioned that neuroprostheses provide a challenge for both moral and legal notions of responsibility (Schermer 2009, Clausen 2013, Burwell et al 2017, Kellmeyer 2018, Jeong et al 2019, Sample et al 2019, Stieglitz 2019, Chari et al 2021 and that algorithms used in neuroprostheses may not be fully controllable, explainable or predictable (Kellmeyer et al 2016, Reilly 2020). Furthermore, researchers should learn about ethics during their education and training in order to ensure responsible innovation (Tracey and Flower 2014).

Involvement of experts, patients and the public
The involvement of experts, patients and the public was argued to be important throughout the development process of neuroprostheses. An overview of 16 implications can be found in supplemental file 12.
First, multidisciplinary teams were argued to be important for successful research, development and the application of neuroprostheses (Vaughan and Wolpaw 2011, Mikołajewski and Mikołajewski 2013, Carmichael and Carmichael 2014, Glannon 2016, Goering and Yuste 2016, Klein 2016, Klein and Ojemann 2016, Moritz et al 2016, Rosenfeld and Wong 2017, Ramos et al 2018, Sullivan et al 2018, Carter et al 2019, Jeong et al 2019, Stieglitz 2019, Adams et al 2020, Chari et al 2021, MacDuffie et al 2021. Second, the involvement of patients and the public was considered important: ethically sound translation of these technologies requires responsiveness to the needs of end-users (Lane et al 2011, Vaughan and Wolpaw 2011, Jebari and Hansson 2013, Xia and Ren 2013, Moritz et al 2016, Rosenfeld and Wong 2017, Kellmeyer 2018, Panuccio et al 2018, Sullivan et al 2018, Ereifej et al 2019, Sample et al 2019. Several articles mentioned that the knowledge and experiences of end-users are useful for the design of a neuroprosthesis device, for the design of a research study, and for informed consent (Lane et al 2011, Weisleder 2012, Moritz et al 2016, Sullivan et al 2018, Sample et al 2019. When patients are actively involved throughout the research process, their involvement has a positive effect on their experience of this research process (Lane et al 2016). Furthermore, researchers should include the perspectives of a group of (potential) end-users consisting of both proponents and opponents of the technology, to prevent selection bias in end-users engagement (Sullivan et al 2018). Empirical (qualitative) research into users' experiences, perceptions and needs with regards to neuroprosthesis usage is hardly ever conducted, resulting in a poor understanding of the bodily experience of those using these technologies (Lane et al 2011, Klein and Ojemann 2016, Quigley and Ayihongbe 2018, Sullivan et al 2018, Kogel et al 2019, Sample et al 2019, Tbalvandany et al 2019. Furthermore, efforts to make research participatory should strike a balance between scientific quality and benefit to participants (Lane et al 2011). Communication with the broader public is also important: socially acceptable design requires the involvement of the public and the development of ways to inform the public about neuroprostheses (Lane et al 2011, Jebari and Hansson 2013, Goering and Yuste 2016, Moritz et al 2016, Burwell et al 2017, Adams et al 2020. To foster realistic expectations of the public, responsible media coverage is needed (Haselager et al 2009, Laryionava and Gross 2011, Xia and Ren 2013, Carmichael and Carmichael 2014, Goering and Yuste 2016, Klein 2016, Moritz et al 2016, Burwell et al 2017, Sullivan et al 2018, Carter et al 2019. The literature also mentioned that the legislative process for neuroprostheses should be made transparent to the public (Jebari and Hansson 2013).
Lastly, the needs of end-users have to be understood in order to develop neuroprostheses that are acceptable to them (Lane et al 2011, Moritz et al 2016, Sullivan et al 2018, Ereifej et al 2019. As neuroprostheses affect society at large, it is important to understand public interests and perceptions of neurotechnology, to determine whether these technologies are socially acceptable and whether the public has a realistic image of these technologies (Vlek et al 2012, Moritz et al 2016, Carter et al 2019, Kellmeyer 2019. Furthermore, awareness of the invasiveness and risks of neuroprostheses are important for acceptance by society (Jebari andHansson 2013, Rosenfeld andWong 2017). Images in movies and the media can have an influence on public attitudes towards neuroprostheses and enablement or further limitation of the technology (Wurzman et al 2017). Furthermore, some authors mentioned that while it remains unclear whether 'cyborgs' will (ever) be accepted by society (Zehr 2015), public perception may improve once the benefits are better understood (Lucas 2012).

Discussion
Despite an ongoing academic debate on the advancing field of neuroprostheses, this study is the first comprehensive overview of the ethical implications of visual neuroprostheses. The rapid technological advancements spark hope for further translation of these technologies to patients and society at large, for which ethical guidance is essential. In what follows, we discuss our findings in the light of existing literature on neurotechnology, reflect on the translation of these devices into society, and formulate an outlook for future research.

The ethics of visual neuroprostheses in the light of other neurotechnologies
Having reviewed the literature, we identified 169 ethical implications both of visual neuroprostheses and of related neurotechnologies such as CIs and BCIs.
Some of these implications are specific to visual neuroprostheses. For example, the literature on visual neuroprotheses showed that complete restoration of visual function is not yet possible, and there are still open questions as to what would be considered meaningful restoration (Lane et al 2011, Xia and Ren 2013, Rosenfeld and Wong 2017, Wood et al 2019, Coin and Dubljević 2020, Hansson 2020. Another example of a specific implication of visual neuroprostheses is their potential to increase the user's independence by enabling navigation through their environment (Lane et al 2012(Lane et al , 2016. Other implications have been identified that are applicable to the broader field of neurotechnologies and are expected to become relevant for visual neuroprostheses, though whether and how exactly still remains to be seen. For example, some ethical implications focus specifically on children, but visual neuroprostheses are not expected to be used in children in the near future. Nevertheless, these implications may become applicable to visual neuroprostheses, if they become available to children in the coming decades. Another implication that remains poorly understood is related to personality change (Schermer 2009, Laryionava and Gross 2011, Lane et al 2012, Jebari and Hansson 2013, Mikołajewski and Mikołajewski 2013, Goering and Yuste 2016, Moritz et al 2016, Mudgal et al 2020. For the visual neuroprosthesis such a change may be caused by clinical side-effects of the implantation of the neuroprosthesis in the brain, for instance due to a structural alteration or bleeding. A second, more indirect, way in which a user's personality or sense of self could be altered is related to the desired restoration of visual function. This restoration of function could result in the so-called Burden of Normality, which is often described in discussions on the adjustment of epilepsy patients to being symptomfree and the duties that come with a symptomfree life after successful (surgical) treatment (Gilbert 2012).
Still other ethical implications are generally applicable to all types of implanted neurotechnologies, for example (surgical/mechanical) risk and harm that are likely to be directly translatable to visual neuroprostheses. (Kotchetkov et  . This is even more important as it is currently not known how users are affected by the explantation of a neuroprothesis (Lane et al 2012, Pass and Graber 2015, Burwell et al 2017, Hansson 2020. A topic that has hardly been described in the literature on visual neuroprostheses, but may become prominent in the ethical guidance of these technologies, is the convergence of neuroprostheses with AI. This convergence opens up new ethical challenges and opportunities, including those of data protection, moral and legal responsibility, and autonomy (Kellmeyer 2019, Rainey et al 2020, Chari et al 2021. While it remains unclear whether AI will in fact be applied to this field and what its role will be, the enthusiasm for medical AI for image-processing indicates the potential benefits of such technologies for processing images in visual neuroprostheses (Pesapane et al 2018).
Though we can indeed learn from the ethical implications of the other neurotechnologies included in our review, it is equally important to acknowledge that relevant differences between these technologies may also exist. The category of visual neuroprostheses alone encompasses several different types of devices, accompanied by their own specific set of ethical implications. For the ethical assessment and guidance of these technologies, it is important to differentiate between the different benefits and risks of these devices, as these differences may lead to different conditions under which these devices are ethically and socially acceptable.

Translating visual prostheses into society
The clinical translation of these devices into society highlights the need to reflect on conditions for societal acceptability. Our review has pointed out the necessity of legal and ethical guidance in order to ensure safety and to protect potential users. The current legal framework focuses strongly on the safety of these devices and the protection of individual users (Regulation ( . Applications of these technologies in nonmedical settings -such as gaming and enhancementremain beyond the scope of these regulations (The Royal Society 2019). Additional legal and ethical guidance is needed to ensure access to and fair distribution of these devices and foster the ethically sound and societally acceptable clinical translation of these devices. Our review indicates that further guidance is needed to ensure that oversight and monitoring boards have no conflicts of interest, due to the small size of the community involved, and that potential conflicts of interest due to the influence of commercial companies can be regulated.
Beyond the need to ensure the safety of and access to such devices, the clinical translation of visual prostheses also requires societal sensitivity and responsiveness. Ethically sound translation of these technologies requires responsiveness to end-users' needs (Jongsma and Bredenoord 2020). Given the relevance of societal perception of the acceptability of neurotechnologies and the impact that these technologies may have on society, public engagement is becoming increasingly important to foster trust and to better understand conditions for societal acceptance. Such societal debate should discuss stigma, the humanmachine relationship, desirable aims, and usage of these devices-including enhancement possibilities, and the fair distribution of these relatively costly technical solutions. Importantly, the societal debate may be hampered by unrealistic images, dystopian phrasings and futuristic narratives in the media. This raises a responsibility for journalists and media platforms to consider their role in the perpetuation of hypothetical scenarios that might eventually result in failed innovations (Sand 2018).

Strengths and limitations
This systematic review provides a comprehensive overview of the ethical implications brought forward in the academic literature on (visual) neuroprostheses. The articles were included after a thorough screening of the academic literature on the topic by two independent reviewers, based on a search strategy that was guided by experienced librarians. Nonetheless, this review has several limitations.
First, this review includes articles on neurotechnology in general, BCIs, and CIs. Some of these implications found in the broader literature may turn out to be only marginally relevant for visual neuroprotheses. Second, we have not snowballed the included literature to extend the search for other relevant papers. Third, a systematic review of this kind always involves reporter bias; a different group of researchers could have selected or grouped the included reasons in a different way. We have minimized bias by thoroughly discussing the ethical implications identified within our research team. Fourth, we did not perform a quality assessment of the included literature beyond the requirement to be published in an academic journal as there is no screening instrument available to assess the quality of normative papers. Finally, it was beyond the scope of this paper to assess the scientific validity of the implications and different applications of neuroprostheses discussed in the included articles.

Conclusion and outlook
Visual neuroprostheses have a broad range of ethical implications. Despite the rapid academic advancement, the ethical implications are rarely discussed in the academic literature. This article provides a systematic review of ethical implications of (visual) neuroprostheses as reported in the academic literature, in which lessons are also drawn from experiences with other neurotechnologies. The review includes six papers that specifically focused on visual neuroprostheses and 78 focused on other neurotechnologies. We have identified numerous ethical implications of (visual) neuroprostheses, which are categorized in seven overarching themes: benefits for health and well-being, harm & risk, autonomy, societal effects, clinical research, regulation & governance and involvement of experts, patients and the public. Our review indicates the need to differentiate between different technologies and their specific implications to assess the conditions for acceptability and the need to develop further regulation and ethical guidance for the translation of these technologies into society.
Further empirical research and normative analysis are needed to keep pace with technological advancements with regard to the development, design and application of visual neuroprostheses. This demands a multidisciplinary approach, as many different stakeholders are involved in the field of visual neuroprostheses. Proactive scrutiny and multidisciplinary collaboration can help to turn these technological developments into successful innovations and can reshape the academic and societal debate to allow more realistic images and expectations of these technologies.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).

Acknowledgments
EAMV was responsible for the original article design, data collection and analysis, drafting and revision for important intellectual content. D R E H and KRJ played a significant role in the article design, data collection and analysis. O C S, D R E H, M B, X C, P R, A L B and K R J were responsible for the concept and design, providing important intellectual input, critically reviewing the drafts and final approval of the article. K R J is assistant professor of Bioethics at the University Medical Centre of Utrecht, and the guarantor for this article. She accepts full responsibility for the work and the conduct of the study. Her involvement was critical to every phase of this work and she controlled the decision to publish. This work is part of the INTENSE consortium, which has received funding from the Dutch Research Council (NWO) with Grant No. 17619.