8 120 8. General discussion investigate task learning effects with SPV (e.g., Dagnelie et al., 2006). Although it was not the focus of the current dissertation, it is relevant to acknowledge the value of SPV studies for the development of effective training and rehabilitation strategies. 8.4.1. Societal and ethical implications There are many societal and ethical considerations related to the development of bionic vision. Although an extensive discussion is outside the scope of this dissertation, it is relevant to note that there is an increasing body of literature on socio-ethical implications of visual neuroprostheses (e.g., seevan Velthoven et al., 2022for a review). Several of the most relevant or related topics are named here briefly. Firstly, it is important to consider questions regarding the risks versus benefits that users are exposed to when obtaining prosthetic vision. How can adequate expectations be ensured of prospective users? How can prospective users be supported to make an informed decision? Besides clinical risks, how do we avoid technological risks, such as brain-hacking? To some extent, these questions sound no different from other risk-benefit assessments in the clinical practice. However, the different nature of prosthetic vision compared to natural sight makes that it is difficult to shape expectations. Both early-stage clinical studies and simulation research can help to better anticipate functional outcomes. Prosthetic engineers will need to take into account digital security of the technology (Ienca, 2015). Further developments in the field of medical engineering, such as the development of flexible electrodes can help to reduce risks and improve the functional quality (Fernández & Botella, 2018). For prospective users, a clear understanding of the potential risks and benefits is crucial for making an informed autonomous decision. Secondly, there are considerations related to the accessibility and social equality. How can treatment with visual prosthetic be made accessible and how should inequality be avoided? What are the risks of losing long-term support? What about software updates? This type of issue is difficult to address in isolation, and it requires cooperation between industry, research, healthcare and government. A strong collaboration between research and industry could help to translate experimental technology into a commercially (and clinically) viable product. Legislation might be necessary to warrant long-term support. Valuable lessons can also be learned from early-stage retinal implants as well as comparable neurotechnology such as cochlear implants (van Velthoven et al., 2022). Thirdly, there are many specific considerations related to the development of visual processing algorithms. To what extent are prosthetics engineers entitled to determine what the user will get to ‘see’? How is it decided what is relevant for the user? To what extent is it permitted to provide super-visual input to the prosthesis user? How cultural or social biases be avoided in the creation of scene processing software? To begin with the last question, it is important to understand - and prevent - the introduction of many different biases in the development artificial intelligent systems. Note that ’algorithmic fairness’ is an active field of study (Mehrabi et al., 2022). But even without social or cultural biases, scene processing algorithms - by design - give a selective representation of the environment, controlling the information input of the user. To increase autonomy, it is considered to offer multiple vision ’modes’ that can be activated by the user (Beyeler & Sanchez-Garcia, 2022; Lozano et al., 2020). The involvement of patients during the development can help to include the priorities of the prospective users. Altogether, a
RkJQdWJsaXNoZXIy MTk4NDMw