Digital Extended Reality could change how people connect with each other and their surroundings in physical and virtual settings.
We include two many technologies in this family: Extended Reality (XR), which relates to virtual and simulated experiences using digital technologies, and Natural Language Processing (NLP), which allows computer systems to process and analyse a vast quantity of human natural language information (e.g., voice, text, images) and generate text in natural or artificial languages. These two technologies can stand alone or be combined in certain devices. You can explore specific examples that fall in these categories below.
Potential ethical repercussions of such technologies include cognitive and physiological impacts as well as behavioural and social dynamics, such as influencing users’ behaviours, and monitoring and supervising people.
Ethical analysis
‘Ethics by design’ is at the core of TechEthos. It was necessary to identify the broad array of values and principles at stake in Digital Extended Reality, to be able to include them from the very beginning of the process of research and development. Based on our ethical analysis, we will propose how to enhance or adjust existing ethical codes, guidelines or frameworks.
XR and NLP are treated as self-sufficient, standalone technologies in the analysis below, but our in-depth reports also look at the ethical issues raised by their combination.
In there a preference for material reality?
The emergence of virtual reality prompts the question of whether virtual experiences mediated via XR are equivalent to experiences gained in the real world: do they evoke similar emotions, behaviours or judgements?
Mode of being of virtual objects
Digital objects are the types of things we experience in the digital world, like “an image” or “a video”. However, it is not clear how they can be individual objects if all they consist of is digital data. The philosophy of digital objecthood features several position. A moderate one is that digital objects exist insofar as they are experienced and conceptualised by a digital mind. A more radical position claims that virtual objects and environments are of the same nature as material objects and environments.
Value of virtual objects
If a distinction between virtual objects and material objects is kept, consequences of actions in material reality certainly do not equal the consequences of actions in virtual reality. For example, driving fast in virtual reality does not imply the same risk as driving fast on a material road.
Nevertheless, some scholars have argued that virtual objects do retain some ethical value, not because of the equivalent consequences involved, but because values or behaviour patterns formed in XR can influence behaviours in the real world, for example in speeding on a road in one’s actual car, with negative consequences.
More core ethical dilemmas are tackled in the ‘Analysis of Ethical Issues’ report.
Training: knowledge transfer and qualia
One of the most established applications of XR is in training skills. The areas in which XR training applications are the most impactful usually include high-risk or costly material training conditions, such as traing for pilots and surgeons. Are skills acquired via virtual experiences equivalent or transferable to material conditions?
Remote work: long-term effects on workers and the job market
XR work environments are available on the market and allow coworkers to host meetings and interact at a distance, sometimes using avatars. The ethical challenges associated with its use include the potential overuse of this always-accessible mode of work, impact on local job markets, and the collection of workers’ data, among others.
More applications and use cases are tackled in the ‘Analysis of Ethical Issues’ report.
NLP systems lack human reasoning
Today most chatbots are deterministic models without machine learning. They take the user down a decision tree in a predetermined way. However, the most advanced NLP techniques, capable of varied conversation on many topics with nearly human-level outputs, rely on statistical linguistic analysis. They do not involve any understanding of meaning or semantics. Void of intention and disconnected from action and responsibility, they cannot be considered on a par with language produced by human speakers. However, humans might take the chatbot’s language to be meaningful and react to its semantic content.
Artificial emotions influence human users
Some applications use conversational agents to influence their users through the architecture or language of the dialogue. Manipulation by a conversational agent can be direct (including inaccurate or skewed information) or indirect, using the “nudging” strategies.
More core ethical dilemmas are tacked in the ‘Analysis of Ethical Issues’ report.
Human resources: gender bias, data protection and labour market
Chatbots are used by human resources managers for recruitment as well as for career follow-up and employee training. The training data used have been found to be biased, especially against margionalised populations. This can tlead to different types of harm, in terms of how these populations are represented and what resources or opportunities, such as jobs, are allocated to them.
Creativity: authenticity
NLP can be used to generate seemingly creative or poetic text that has no human creative input or that relies on prior creative work. If such applications were used at scale, it might reduce the profitability of creative or innovative work.
More applications and use cases are tackled in the ‘Analysis of Ethical Issues’ report.
