180
Chapter 11: Artificial Intelligence
his or her point of view and to contest the decision.
For guidance related to automated-decision making
and profiling under GDPR, see wp251rev.01 published
by Art. 29 Working Part (WP29), re-endorsed by
European Data Protection Board (EDPB) as GDPR:
Guidelines, Recommendations, Best Practices. EDPB is the
successor of WP29. EDPB is the entity comprising the
national information protection authorities. It guards
and enforces the consistent implementation of the
GDPR across the EU. European Commission website.
https://ec.europa.eu/newsroom/article29/item-detail.
cfm?item_id=612053. Accessed 16 February 2021.
49. There are many different interpretations of fairness. The
European Parliament Panel for the Future Science and
Technology (STOA) in its study, Artificial intelligence:
From ethics to policy (June 2020), defines fairness as
the requirement to equally distribute goods, wealth,
harms, and risks. The General Data Protection Regulation
(GDPR) refers to substantive fairness (Recital 71) as
fairness of the content of an automated inference or
decision, according to the STOA report. The impact of
the GDPR on artificial intelligence (25 June 2020) can
be summarized as (a) Acceptability, i.e., the input data
(the predictors) for the AI decision being normatively
acceptable as a basis for the inferences concerning
individuals (e.g., the exclusion of ethnicity if this does
not impact disease determination), (b) Relevance: the
inferred information (the target) should be relevant to
the purpose of the decision and normatively acceptable
in that connection, (c) Reliability: both input data,
including the training set, and the methods to process
them should be accurate and statistically reliable
(which is an aspect that must be proven in the clinical
evaluation or performance evaluation of a medical
device). On the other hand, the GDPR also refers to
informational fairness (Recital (60): that the subject
must be informed of the existence of the processing
operation and its purposes. The High-Level Expert
Group on Artificial Intelligence considers fairness to
have both a substantive and a procedural dimension.
The substantive dimension implies a commitment to
ensuring equal and just distribution of both benefits
and costs and ensuring that individuals and groups are
free from unfair bias, discrimination, and stigmatization.
The procedural dimension of fairness entails the ability
to contest and seek effective redress against decisions
made by AI systems and by the humans operating
them. Ethics Guidelines for Trustworthy AI, 8 April
2019. The European Parliament Panel for the Future of
Science and Technology (STOA) defines fairness as the
requirement to equally distribute goods, wealth, harms,
and risks. Artificial Intelligence: From Ethics to policy.
24 June 2020. European Parliament website. https://
www.europarl.europa.eu/stoa/en/document/EPRS_
STU(2020)641507. Accessed 16 February 2021.
50. Artificial intelligence: From ethics to policy. European
Parliament Panel for the Future of Science and
Technology (STOA). June 2020. European Parliament
website. https://www.europarl.europa.eu/stoa/en/
document/EPRS_STU(2020)641507. Accessed 16
February 2021.
51. Winfield A, et. al. The Ethics of Artificial Intelligence:
Issues and initiatives. Panel for the Future of Science
and Technology (STOA) Panel. 11 March 2020.
Science Communication Unit at the University of the
West of England at the request of STOA. European
Parliament website. https://www.europarl.europa.eu/
stoa/en/document/EPRS_STU(2020)634452. Accessed
16 February 2021.
52. Op cit 44.
53. Op cit 50.
54. Lebeer G. Ethical Function in Hospital Ethics
Committees. IOS Press, 2002.
55. A statistic and a parameter are very similar. They are
both descriptions of groups. The difference between
a statistic and a parameter is that statistics describe a
sample. A parameter describes an entire population.
A statistic is biased if it is calculated in such a way
that it is systematically different from the population
parameter being estimated.
56. Article 4 (m) of draft Report with recommendations to
the Commission on a framework of ethical aspects of
artificial intelligence, robotics and related technologies
(2020/2012(INL)). 4 April 2020. Committee on
Legal Affairs. European Parliament website. https://
www.europarl.europa.eu/doceo/document/JURI-
PR-650508_EN.html?redirect Accessed 16 February
2021.
57. The scientific definition of bias uses the words
“tendency” and “statistic,” which alludes to there being a
systematic nature.
58. Women, because it is harder to find a woman that is
not childbearing to participate in clinical investigations
minority racial or ethnic groups because it is harder
to find enough people of that subgroup elderly,
because it sometimes carries more risks to include
them. Fox-Rawlings SR, Gottschalk LB, Doamekpor
LA, Zuckerman DM, and Milbank Q. Diversity in
Medical Device Clinical Trials: Do We Know What
Works for Which Patients? 2018 96(3):499–529.
doi:10.1111/1468-0009.12344.
59. Hébert-Johnson U, Kim MP, Reingold O,
Rothblum GN. Multicalibration: Calibration for
the (Computationally-Identifiable) Masses.” Cornell
University website. https://arxiv.org/abs/1711.08513.
Accessed 16 February 2021.
60. Jobin A, et.al. Artificial Intelligence: the global
landscape of ethics guidelines. June 2019.
ResearchGate website. https://www.researchgate.net/
publication/334082218_Artificial_Intelligence_the_
global_landscape_of_ethics_guidelines. Accessed 16
February 2021.
61. High-Level Expert Group on AI (2020). Ethics
Guidelines for Trustworthy AI.
Chapter 11: Artificial Intelligence
his or her point of view and to contest the decision.
For guidance related to automated-decision making
and profiling under GDPR, see wp251rev.01 published
by Art. 29 Working Part (WP29), re-endorsed by
European Data Protection Board (EDPB) as GDPR:
Guidelines, Recommendations, Best Practices. EDPB is the
successor of WP29. EDPB is the entity comprising the
national information protection authorities. It guards
and enforces the consistent implementation of the
GDPR across the EU. European Commission website.
https://ec.europa.eu/newsroom/article29/item-detail.
cfm?item_id=612053. Accessed 16 February 2021.
49. There are many different interpretations of fairness. The
European Parliament Panel for the Future Science and
Technology (STOA) in its study, Artificial intelligence:
From ethics to policy (June 2020), defines fairness as
the requirement to equally distribute goods, wealth,
harms, and risks. The General Data Protection Regulation
(GDPR) refers to substantive fairness (Recital 71) as
fairness of the content of an automated inference or
decision, according to the STOA report. The impact of
the GDPR on artificial intelligence (25 June 2020) can
be summarized as (a) Acceptability, i.e., the input data
(the predictors) for the AI decision being normatively
acceptable as a basis for the inferences concerning
individuals (e.g., the exclusion of ethnicity if this does
not impact disease determination), (b) Relevance: the
inferred information (the target) should be relevant to
the purpose of the decision and normatively acceptable
in that connection, (c) Reliability: both input data,
including the training set, and the methods to process
them should be accurate and statistically reliable
(which is an aspect that must be proven in the clinical
evaluation or performance evaluation of a medical
device). On the other hand, the GDPR also refers to
informational fairness (Recital (60): that the subject
must be informed of the existence of the processing
operation and its purposes. The High-Level Expert
Group on Artificial Intelligence considers fairness to
have both a substantive and a procedural dimension.
The substantive dimension implies a commitment to
ensuring equal and just distribution of both benefits
and costs and ensuring that individuals and groups are
free from unfair bias, discrimination, and stigmatization.
The procedural dimension of fairness entails the ability
to contest and seek effective redress against decisions
made by AI systems and by the humans operating
them. Ethics Guidelines for Trustworthy AI, 8 April
2019. The European Parliament Panel for the Future of
Science and Technology (STOA) defines fairness as the
requirement to equally distribute goods, wealth, harms,
and risks. Artificial Intelligence: From Ethics to policy.
24 June 2020. European Parliament website. https://
www.europarl.europa.eu/stoa/en/document/EPRS_
STU(2020)641507. Accessed 16 February 2021.
50. Artificial intelligence: From ethics to policy. European
Parliament Panel for the Future of Science and
Technology (STOA). June 2020. European Parliament
website. https://www.europarl.europa.eu/stoa/en/
document/EPRS_STU(2020)641507. Accessed 16
February 2021.
51. Winfield A, et. al. The Ethics of Artificial Intelligence:
Issues and initiatives. Panel for the Future of Science
and Technology (STOA) Panel. 11 March 2020.
Science Communication Unit at the University of the
West of England at the request of STOA. European
Parliament website. https://www.europarl.europa.eu/
stoa/en/document/EPRS_STU(2020)634452. Accessed
16 February 2021.
52. Op cit 44.
53. Op cit 50.
54. Lebeer G. Ethical Function in Hospital Ethics
Committees. IOS Press, 2002.
55. A statistic and a parameter are very similar. They are
both descriptions of groups. The difference between
a statistic and a parameter is that statistics describe a
sample. A parameter describes an entire population.
A statistic is biased if it is calculated in such a way
that it is systematically different from the population
parameter being estimated.
56. Article 4 (m) of draft Report with recommendations to
the Commission on a framework of ethical aspects of
artificial intelligence, robotics and related technologies
(2020/2012(INL)). 4 April 2020. Committee on
Legal Affairs. European Parliament website. https://
www.europarl.europa.eu/doceo/document/JURI-
PR-650508_EN.html?redirect Accessed 16 February
2021.
57. The scientific definition of bias uses the words
“tendency” and “statistic,” which alludes to there being a
systematic nature.
58. Women, because it is harder to find a woman that is
not childbearing to participate in clinical investigations
minority racial or ethnic groups because it is harder
to find enough people of that subgroup elderly,
because it sometimes carries more risks to include
them. Fox-Rawlings SR, Gottschalk LB, Doamekpor
LA, Zuckerman DM, and Milbank Q. Diversity in
Medical Device Clinical Trials: Do We Know What
Works for Which Patients? 2018 96(3):499–529.
doi:10.1111/1468-0009.12344.
59. Hébert-Johnson U, Kim MP, Reingold O,
Rothblum GN. Multicalibration: Calibration for
the (Computationally-Identifiable) Masses.” Cornell
University website. https://arxiv.org/abs/1711.08513.
Accessed 16 February 2021.
60. Jobin A, et.al. Artificial Intelligence: the global
landscape of ethics guidelines. June 2019.
ResearchGate website. https://www.researchgate.net/
publication/334082218_Artificial_Intelligence_the_
global_landscape_of_ethics_guidelines. Accessed 16
February 2021.
61. High-Level Expert Group on AI (2020). Ethics
Guidelines for Trustworthy AI.