A brief summary of our round table in London
On the 11th of June, we celebrated in the impressive facilities of Allen & Overy the first round-table on the Code of Conduct code (see invite)
We welcomed artificial intelligence leaders, data privacy experts, analytics thought leaders, etc. Our Allen & Overy was represented by the partners Nigel Parker and Jane Finlayson-Brown (specialized in intellectual property, data protection and privacy).
From PriceWaterhouseCoopers, we welcomed Euan Cameron (AI Leader UK) and Neil Hampson (UK Data Analytics Leader). Tim McGarr (Market Development Manager Digital ) joined from the British Standards Institution. Koichi GOTO (Business Development for Japanese and South East Asian) brought us very interesting perspectives from Japan, a country where Robotics and AI are very much embedded in the society and industry.
John Barker joined from the Alan Turing Institute Commercial Development Board.
We opened up the round explaining the motivations behind the code of conduct. Peter Grindrod… rather than “Deep-sea ethical fishing” approaches, with the aim of addressing the challenges the mankind will face in the mid-to-long term according to the projected development of artificial intelligence technologies, our code focuses on here and now… on the state of the art and on the existing challenges of the data science today.
The participants unanimously welcomed the focus and the approach and expressed their support.
After that, I run the audience through the different sections of the code. We had the opportunity to thoroughly discuss each and every clause and at the end, each participant shared her/his feedback:
- On “The Data Scientist should not misrepresent his/her past experiences, past achievements, domain authority, or educational qualifications.”
The participants discussed the need for accredited data science degree as well as the definition of data science. A common domain Taxonomy is required (“what do we mean by data scientists? what is excluded from the definition?
- On “Sharing of Lab Log”
Potential conflicts identified between transparency and IP Protection. Also the way of sharing information needs to be fixed and documented, when it is implemented.
- On “Accuracy importance depending on the nature of the problem” and all the mentions for “accuracy”
The participants well pointed out that “accuracy” might be too narrow, as additional quality measures might be more relevant depending on the nature of the problem. As “add-on” they suggested the responsability of selecting the right metric/KPI for the right problem.
- On “Liability in case of failure of the Data Scientist’s model”.
It’s critical to educate algorithm users (usually lacking data science background) in the hand-off process. Technical “Terms and Conditions” are probably not sufficient
- The code is defined from the data scientist’s perspective. What would be the perspective of an employer?
Ethical culture shall be enforced by the employer. Ideally, employers can show compliance
- Companies shall provide a mechanism to enable the data scientists to speak up (protected disclosures / whistle-blowing)
We all agreed on incorporating the feedback to the code and to organize a kick-off event with the top 30 UK companies in London after the round table in Munich has taken place (September).
In addition, Peter and myself will coordinate the creation of a new, more compact version of the code and make it available in other languages.