By Sara R. Jordan, member, IEEE Global Initiative for Ethical Considerations in Artificial Intelligence and Autonomous Systems;
Assistant Professor in the Center for Public Administration and Policy at Virginia Tech
Bringing Ethics into
Engineers have two established models to investigate and implement ethics in product design decisions.
Day after day, major news outlets offer readers a juicy story about the promises or perils of autonomous
systems. Lost in the melee of chatter about a robo-apocalypse, is the hard-nosed reality of designing
autonomous systems (AS) to make decisions that meet
our ethical and aesthetic preferences, while also meeting
good engineering design specifications. Designing ethical
products isn’t a problem confined only to robotics,
machine learning experts, and components engineers.
Those working in the AS space have special challenges
and responsibilities for designing ethics into their systems.
The state of the engineering field is changing
to incorporate end-user’s aesthetic and ethical
considerations, along with well-known bodies in the
engineering profession that have taken up the task of
helping practicing engineers to identify, prioritize, choose,
and embed ethical values into AI/AS designs. Building
on the long-lasting initiatives of accrediting bodies
such as ABET, professional associations like the NSPE
academic researchers such as those publishing in journals
like Science & Engineering Ethics or IEEE Spectrum, and
leading industry practices, such as consideration of values
in agile project management, ethical analysis is now a
normal part of 21st century engineering practice.
Regardless of educational trends, engineers aren’t going
to turn into professional ethicists any time soon, nor
should they be expected to become professional ones.
Instead, practicing engineers need tools to meet these
evolving demands for ethics in their design spaces. What
tools are available to engineers who need to address
ethical concerns during any stage of the product lifecycle?
In broad strokes, there are two competing models
for investigating ethics during product design. One,
value sensitive design (VSD), incorporates ethical
considerations using an iterative model of values guiding
product design. Under the VSD model, ethical principles
that design teams believe they must adhere to emerge
within the process of design itself. Another, the ethical
decision matrix approach, adapts the tools of decision-matrices, according to which proposed designs are
debated in relation to ex ante specified value parameters.
Values by Design
VSD incorporates intuitive, empirical, qualitative, and
quantitative, approaches to reasoning through the design
of software algorithms, and hardware systems. There
are four stages of value-sensitive design including: value
discovery, conceptualization, empirical value investigation,
and technical value investigation (Spiekerman 2015 p.
168). Each stage builds iteratively upon the discoveries
made in the previous stage, allowing for values to emerge
throughout the process.
Value discovery is a component of good product design
that engineers already do, and for which they have well-developed tools. For example, cost benefit analysis or
threat identification are one form of values discovery tool
that help design teams to identify who and what they
care about most, such as environmental sustainability
or equality of access for disabled users. While harm
reduction is an almost universal value for design teams
to strive, new value expressions, such as human well-being, are being advanced as state of the art for ethical
engineering practice in the AI/AS space.
According to authors of Ethically Aligned Design: A
Vision for Prioritizing Human Wellbeing with Artificial
Intelligence and Autonomous Systems (EAD, Version 1)
and the authors of the Ethical Design Manifesto, human
well-being is a paramount value for teams to strive in their
designs. Far from being merely a measure of wealth or
purchasing power, the authors of a well-being approach to
design ethics suggest that teams take into consideration
new, qualitative measures like the Better Life Index or the
United Nations Sustainable Development Goals.
Well-being and other high-level values, such as social
justice, may be hotly debated during the values discovery
process. The debate about ultimate values and their
meaning is not ‘merely academic’ as it guides discussion
about implementation of those values. This second stage
of the design process—values identification—prompts
engineering teams to discuss the tensions that exist
between technological specifications and ethical goals.
Importantly for components engineers, identifying the
priorities of values held by different component design