Why AI ethics?

Why AI ethics?

Importance of AI ethics has largely emerged as a response to the range of individual and societal harms that the misuse, abuse, poor design, or negative unintended consequences of AI systems may cause. The importance of building a robust culture of AI ethics within an organization must not be taken lightly as AI technology has emerged as the most transformative and disruptive technology every adoption may have enormous benefits but also unpredictable negative consequences and potential harms.

Potential Harms Caused by AI Systems

1. Bias and Discrimination

Because they gain their insights from the existing structures and dynamics of the societies they analyze, data-driven technologies can reproduce, reinforce, and amplify the patterns of marginalization, inequality, and discrimination that exist in these societies.

Likewise, because many of the features, metrics, and analytic structures of the models that enable data mining are chosen by their designers, these technologies can potentially replicate their designers’ preconceptions and biases.

Finally, the data samples used to train and test algorithmic systems can often be insufficiently representative of the populations from which they are drawing inferences. This creates real possibilities of biased and discriminatory outcomes, because the data being fed into the systems is flawed from the start.

2. Denial of Individual Autonomy, Recourse, and Rights

When citizens are subject to decisions, predictions, or classifications produced by AI systems, situations may arise where such individuals are unable to hold directly accountable the parties responsible for these outcomes. AI systems automate cognitive functions that were previously attributable exclusively to accountable human agents.

This can complicate the designation of responsibility in algorithmically generated outcomes, because the complex and distributed character of the design, production, and implementation processes of AI systems may make it difficult to pinpoint accountable parties.

In cases of injury or negative consequence, such an accountability gap may harm the autonomy and violate the rights of the affected individuals.

3. Non-transparent, Unexplainable, or Unjustifiable Outcomes

Many machine learning models generate their results by operating on high dimensional correlations that are beyond the interpretive capabilities of human scale reasoning. In these cases, the rationale of algorithmically produced outcomes that directly affect decision subjects remains opaque to those subjects.

While in some use cases, this lack of explainability may be acceptable, in some applications, where the processed data could Understanding Artificial Intelligence Ethics and Safety 5 harbor traces of discrimination, bias, inequity, or unfairness, the opaqueness of the model may be deeply problematic.

4. Invasions of Privacy

Threats to privacy are posed by AI systems both as a result of their design and development processes, and as a result of their deployment. As AI projects are anchored in the structuring and processing of data, the development of AI technologies will frequently involve the utilization of personal data. This data is sometimes captured and extracted without gaining the proper consent of the data subject or is handled in a way that reveals (or places under risk the revelation of) personal information.

On the deployment end, AI systems that target, profile, or nudge data subjects without their knowledge or consent could in some circumstances be interpreted as infringing upon their ability to lead a private life in which they are able to intentionally manage the transformative effects of the technologies that influence and shape their development. This sort of privacy invasion can consequently harm a person’s more basic right to pursue their goals and life plans free from unchosen influence.

5. Isolation and Disintegration of Social Connection

While the capacity of AI systems to curate individual experiences and to personalise digital services holds the promise of vastly improving consumer life and service delivery, this benefit also comes with potential risks. Excessive automation, for example, might reduce the need for human-to-human interaction, while algorithmically enabled hyper-personalisation, by limiting our exposure to worldviews different from ours, might polarise social relationships. Well-ordered and cohesive societies are built on relations of trust, empathy, and mutual understanding. As AI technologies become more prevalent, it is important that these relations be preserved.

6. Unreliable, Unsafe, or Poor-Quality Outcomes

Irresponsible data management, negligent design and production processes, and questionable deployment practices can, each in their own ways, lead to the implementation and distribution of AI systems that produce unreliable, unsafe, or poor-quality outcomes. These outcomes can do direct damage to the wellbeing of individual persons and the public welfare. They can also undermine public trust in the responsible use of societally beneficial AI technologies, and they can create harmful inefficiencies by virtue of the dedication of limited public resources to inefficient or even detrimental AI technologies.