Same AI + Different Deployment Plans = Different Ethics

By neub9
5 Min Read

This month, I would like to delve into an aspect of the ethics surrounding artificial intelligence (AI) and analytics that often goes unnoticed. The ethical implications of an algorithm can vary depending on the specific scope and context in which it is being implemented. What may be considered unethical in one situation may be entirely acceptable in another. I will illustrate this point with an example and then provide guidelines on how to ensure the ethical deployment of AI.

Why Autonomous Cars Aren’t Ethical For Broad Deployment Yet

Currently, there are limited trials of fully autonomous, driverless cars taking place in various parts of the world. However, these cars are primarily operating on low-speed city streets where they can come to a quick stop if necessary. Even in these controlled environments, issues have arisen. For instance, there have been incidents of autonomous cars stopping unnecessarily, causing traffic congestion.

We have not yet witnessed fully autonomous cars operating at high speeds on complex roadways. The main reason for this is the increased risk associated with higher speeds and unpredictable traffic conditions. If an autonomous car encounters a situation it cannot handle while moving at 15 miles per hour, stopping abruptly is a safe option. However, the same scenario at 65 miles per hour could result in a catastrophic accident. Until we are confident that autonomous cars can navigate every possible scenario safely, including novel ones, it is not ethical to unleash them on a large scale on public roads.

Some Massive Vehicles Are Already Fully Autonomous – And Ethical!

If fully autonomous cars are not ethically viable yet, one might assume that huge farm equipment with spinning blades and significant size would be even riskier. However, this is not the case. Manufacturers like John Deere have developed fully autonomous farm equipment that operates in fields autonomously. Despite its massive and potentially hazardous nature, this equipment is considered ethical. Why?

In this scenario, the autonomous farm equipment operates in an isolated field at a slow pace with few obstacles. If the equipment encounters a situation it cannot handle, it stops and alerts the farmer via an app. The farmer can then assess the situation through images sent by the equipment and provide instructions accordingly. This partnership between the autonomous equipment and the human farmer ensures ethical deployment as the risks are minimized and managed effectively.

Implementing the Scope And Context Concept

From the example above, several key takeaways emerge. Firstly, labeling an AI algorithm or application as solely ethical or unethical is overly simplistic. It is crucial to evaluate the specific scope and context of each deployment individually to determine its ethical implications.

Secondly, regular reassessment of past decisions is necessary. As technology advances and circumstances change, the ethical boundaries of AI deployments may shift. Corporate governance and legal regulations may also impact the ethical status of certain deployments. Ethical decisions are time-sensitive and must be revisited periodically.

Lastly, thorough research and consideration of all risks and mitigations are essential. Initial assumptions about the ethical implications of a deployment may not hold true upon closer inspection. Therefore, a comprehensive understanding of the risks involved is imperative to make informed ethical decisions.

Ensuring ethical AI and analytical deployments requires continuous effort and vigilance. Each deployment must be evaluated in its specific context, considering all potential risks and benefits. This underscores the importance of consciously incorporating ethical considerations at every stage of planning, development, and deployment of AI processes, as I have previously emphasized here.

Originally published in the Analytics Matters newsletter on LinkedIn

The post Same AI + Different Deployment Plans = Different Ethics appeared first on Datafloq.

Share This Article
Leave a comment

Leave a Reply

Your email address will not be published. Required fields are marked *