Little robots

Intelligence, Personality & Morality - Defining New Machine Intelligence Quality Characteristics

In our first blog in this series on Machine Intelligence Quality Characteristics, Global Marketing & Communications Director, Paul Saunders looked at the business benefits of creating systematic methods to relate the software quality characteristics of an Artificial Intelligence system to its architecture. In this second blog, AI experts Rik Marselis and Humayun Shaukat, the authors of the Sogeti Report “Machine Intelligence Quality Characteristics - How to Measure the Quality of Artificial Intelligence and Robotics”, describe the new Machine Intelligence Quality Characteristics required for establishing the quality of an AI product. You can download their Report for free here.

The Business Case for New MI Characteristics

Defining and implementing the right MI Quality Characteristics at the design stage, reduces risk, cost, and effort, by enabling quantitative evaluation, prediction and unbiased decision-making about design and design trade-offs. It also enables focused, automated and prioritised quality assurance testing. When done properly, the result is an enhanced AI product with an outstanding user experience, that meets customer and business requirements.

Traditional v’s New Quality Characteristics

There are two categories of Quality Characteristics: Functional, which determine how the end product or service meets end users’ functional requirements, and non-functional which pertain to how these products or services will deliver the functionality.

In our Report we’ve extended the “conventional” Quality Characteristics found in ISO25010, ISO 9126 and Sogeti’s own TMap® NEXT list, to include the characteristics that our research shows are most applicable to AI and Robotics. We’ve added 3 new categories of Machine Intelligence Quality Characteristics: Intelligent Behaviour, Personality and Morality; and also added a new sub-characteristic – Embodiment - to the conventional characteristic of Usability.

Intelligent Behaviour

Intelligent Behaviour can be simply defined as a combination of cognitive skills and knowledge that form an AI system’s ability to understand what is required of it. Intelligent Behaviour includes co-dependent features such as reasoning, memory, imagination, and judgement. These culminate in the ability to learn through rule-based learning, interpretation, observation, and imitation. It’s also essential for humans designing and interacting with these intelligent machines to have total transparency into how they make decisions. Collaboration and natural human interaction such as language and gestures are the final two desirable qualities that make up Intelligent Behaviour.

Morality & the Laws of Robotics

As we reveal in the Report, the Quality Characteristic of Morality is not a million miles away from Isaac Asimov’s Laws of Robotics and is centered around ensuring robots don’t cause human’s any harm, through action or inaction. It is also tied to the concepts of ethics, legality, responsibility, accountability and privacy. This brings into question a lot of interesting and important ideas at the crux of human existence, such as sexuality, love and warfare, all of which we examine in the Machine Intelligence Report in more detail.

The Likeability Factor

We know that we can give machines personality and make them seem humorous, and the way they look and behave can give them charisma and make them loveable. However, can we really instil them with a simulation that even approaches human empathy? Furthermore, how human do we want them to be? Do we want them always to be in the same mood or should we give them the ability to change mood depending on environment and circumstance? Personality is an important quality characteristic that greatly affects the user experience, and which will go a long way to determining whether people love an AI system or find it irritating and frustrating.

Usability & Embodiment

The final new quality sub-characteristic that we address in our Report is “Embodiment”, which comes under the existing category of Usability in the traditional lists of characteristics. What will your robot or system look like? Will it take on a human form or be in more of a charming R2D2 form? How does its physical appearance compare to its capabilities?

To find out the answers to these questions and more, download our Report for free here.

Rik Marselis
Rik Marselis
PointZERO expert / Management Consultant Quality & Testing
+ 31 6 55 69 72 39