AI Ethics – The Fine Line Between Assistance and Autonomy

NDDCEL gives business leaders an invaluable opportunity to expand their understanding of this rapidly advancing topic and make connections with like-minded peers.

Create an AI that is ethical requires more than simply adhering to certain ethical principles; it must also ensure the data used for training purposes is relatively unbiased. Furthermore, privacy must also be addressed.

Ethics and AI

AI’s potential to take jobs away, replace people in medical care and subvert legal structures is real and frightening. While such activities might not always carry clear legal ramifications, they often present us with moral dilemmas which force us to consider accountability not from AI itself but from those who build and use it.

Researching what constitutes ethically acceptable AI can be an exhilarating endeavour spanning linguistics, philosophy, law, psychology, economics and anthropology. For instance, one field of inquiry involves exploring how algorithmic technologies and sociotechnical practices impact individuals’ capacities for practical autonomy (such as their ability to track their identity). Such systems may act either as obstacles to autonomy or facilitate it according to particular values and definitions of autonomy; raising questions as to whether seemingly neutral demands for respecting autonomy actually convey value-laden norms which “ought-to-be”.

Ethics and Machine Learning

Establishing AI ethics requires conducting an exhaustive investigation of major social issues. Enterprises must develop frameworks to assist product teams and executives in designing systems with appropriate ethical considerations in mind.

Non-moral agents do not have duties and cannot directly respect or violate autonomy; however, their behavior can still be held to account through ought-to-do norms. For example, recommendation systems and personalization services that fail to offer meaningful alternatives for self-identification (e.g. non-binary gender categories) could violate an individual’s informational autonomy.

Autonomy requires access to material and, in financialized societies, economic resources. From this viewpoint, one should examine how algorithmic technologies mediate opportunities and distribute resources within an economy, and whether this access promotes or hinders people’s capacity for practical autonomy. Furthermore, considering unintended bias in machine learning algorithms requires continuous detection and tracking to monitor any unintentional bias that arises as mediation takes place – something machine learning algorithms do not do consistently enough.

Ethics and Deep Learning

As AI quickly transforms our world, ethical safeguards must be in place. Otherwise, we risk recreating real-world biases and discriminations, increasing divisions within communities, and undermining fundamental rights and freedoms.

Even weak AI present in algorithms influencing our online shopping recommendations has ethical repercussions due to their propensity to reproduce gender, racial, and cultural biases. This phenomenon can be explained using the classic phrase of ‘garbage in garbage out’ which states that any program or algorithm will reflect values and ideas encoded into it.

Strong AI systems with black box algorithms that defy analysis present even more ethical concerns, as these cannot be considered moral agents or bearers of duties; nor can they be expected to respect or disrespect human autonomy directly. But such AI could still abide by ought-to-be norms: for instance, an AI system could be seen to respect autonomy by not coercing or manipulating people in any way.

Ethics and Robotics

Roboethics is an emerging field of ethics dedicated to robots and artificial intelligence. It primarily deals with how humans should develop, construct, use and treat these systems while also looking at how these systems themselves behave.

This includes how effectively the system allows people to exercise practical autonomy–whether or not it impedes or supports it–in ways that matter, and also what specific scenarios require of the system in terms of what it allows or disallows; such actions must take into account any effects they might have on individual autonomy, dignity and welfare of potential victims.

Many of the same considerations that apply to human ethical agents can also apply to AI agents, and those considerations can often be reduced down into a list of ought-to-be norms based on autonomy (for example protecting vulnerable people and meeting human needs). However, certain issues require new policies to address them effectively — specifically product safety, liability and non-deception advertising practices.