Transparency, fairness and sustainability – these are just a few of the items being called for – also by us – in the creation and use of algorithms.
Those who commission, design and use algorithmic systems – algorithmists – have a responsibility to meet these demands. But what might compel them to fulfill their responsibility? We believe that binding professional ethics can prove effective in this regard.
In our new discussion paper, we take a closer look at the prospects for establishing such ethics. The authors identify ten factors contributing to the success of binding professional ethics in different vocational areas and offer recommendations for applying these practices to the field of algorithm design.
Professional ethics for algorithmists! Calls for ethics in algorithms have been growing louder in recent months – and for good reason, as experts across the board are drawing attention to the ever-growing impact of algorithmic systems on society. And as is now widely understood, humans are responsble for whether or not they do society harm or good. From setting the goals of a system to developing it to selecting training data to system implementation – human beings play a key role as decision-makers.
While this view may be widely acknowledged, in all practicality, there are few ethical demands placed on the professional activity of algorithmists – other than those derived from our broader legal framework. In the English-speaking world in particular, various esteemed institutions have sought to change this by publishing a series of promising quality criteria for algorithms and guidelines for professional ethics in their design.
Bindingness, and how others ensure it
Despite their individual and collective potential, many of these documents lack one thing in particular: bindingness. Without it, the most important demands remain toothless and efforts to to ensure them little more than an academic exercise. For more on the strengths and weaknesses of existing compendia, see here.
Some occupations such as medicine or journalism, for example, have proven successfull in establishing a professional ethic. What are the factors contributing to their binding nature and success? And how can we establish the extent of their success? What can we learn from examples such as the Press Code or Hippocratic Oath? And how might we transfer their practices to the field of algorithm design? These are the questions addressed in our study “Developing a Professional Ethics for Algorithmists: Learning from the Examples of Established Ethics.”
Six different professional ethics and their secrets to success
In the study, Alexander Filipovic, Christopher Koska and Claudia Paganini explore the nature of professional ethics in six different occupational areas: medicine, social work, journalism, public relations, advertising and engineering. The authors examine each area to determine what renders their professional ethics successful and identify a total of ten success factors. These factors are summarized in the figure below. Success in no way contingent on all ten conditions being met. In fact, the absence of some factors can be compensated for by the presence of others.
Lessons for algorithm design
In order to determine the extent to which any factor is transferable to the field of algorithm design, the field itself is first examined more closely in order to define its specific characteristics and differentiate it from other areas more clearly. From the historical roots of algorithm design to the increasing delimitation of the field, the authors draw a broad overview of its features.
Finally, the authors show how the lessons derived from the examples of other professional ethics can be transferred and applied to work on algorithmic systems. They recommend, for example, tracking the historical roots of algorithm creation and the “shedding light on the ethos of those who first created algorithms” as a means of creating a shared sense of historical tradition.
The authors also emphasize the importance of developing and promoting an individual awareness of social and ethical issues among those responsible for the design of algorithms and their implementation. Existing initiatives that address the challenges inherent to establishing an ethics for algorithm design should therefore find support among professionals in the field. Indeed, participatory models of development will foster the personal commitment of individual stakeholders to such ethical standards. These and additional findings are available in our discussion paper. We are also integrating these findings in our #algorules-Process.