IEEE Technology and Society Magazine - December 2019 - 61

Computer-based prediction systems heavily rely on
the outputs of standard mathematical tools to make decisions. This makes the decisions highly quantitative in
nature. Human prediction, on the contrary, relies on various variables such as moods, emotions, biases, etc.,
thus making it qualitative in nature. The fact that computer-based prediction systems are quantitative leads to the
problem of marginalization, whereupon in some mathematical calculations the system marginalizes some data
and while making decisions, it could completely ignore
this marginalized data. This effect has already been demonstrated in photograph classification [34].
This poses some limitations in the decision that the
system makes as there could exist scenarios where
such marginalized data could have played a major role
in decision making. Thus, the rigidity of computer-based
prediction systems, relying only on numbers to make
decisions, could prove futile in such scenarios. On the
contrary, human-based prediction has an ability to
account for this and be "considerate" to such outliers,
probably by ignoring some aspects of the rules and
showing some sort of leniency, while making decisions
in the aforementioned scenarios.
While it is the case that human prediction is influenced by biases, so are computer-based prediction systems. These systems get their biases primarily from the
programmers/scientists who design these algorithms
and systems. However, these biases are constant
unlike, human biases which are random and heavily
influenced by external factors. Hence it would be rather
more feasible to correct an algorithmic bias than a
human bias. This would make the system more robust
in taking decisions with ambiguous rules like the offside
rule in football.
Taking decisions in football for rules like the offside
rule requires rational decision making. A computerbased predictive system makes its decisions based on
numbers and mathematical analysis. A human would
make decisions based on visuals/perception and external factors. This increases the degree of variability in
the decision. A configurable rule-based system can be
trained to learn and improve accuracy of decision making. These decisions can assist human referees and in
effect can improve quality of on-field decisions.
However, there remains a question over whether marginalizing human cognition in decision-making processes
results in a de-skilling of an essential cognitive faculty
[35]. This is most clearly observed in the consideration of
failure. It is not uncommon for the human-only system to
fail, for example for the referee to be injured, but generally the human components are replaceable with like-forlike components. However, if SmartFlag fails and the
fall-back is still to human components, there will be problems if the reliance on automated detection has resulted
DECEMBER 2019

∕

in there being no one who understands or can apply the
rule. At a time when the need for critical thinking skills
has never been more necessary [36], [37] and this skill is
constantly being undermined by developments in Artificial Intelligence, of which SmartFlag could be, quite literally, a flag-bearer, this question needs to be addressed
with some urgency.

Summary and Conclusions
This paper has been concerned with the role of technology in sport, and specifically in the context of Association Soccer, and has introduced a system called
SmartFlag, designed both to detect offside situations
and to decide offside offenses, in real time (by which is
meant: sufficiently fast to be fit-for-purpose).
Some recent events have shown that some technology, such as VAR, only offer limited decision support
because it only offers more detailed review of a situation that already requires interpretation, with respect to
rules that are ever more arcane in their specification
(e.g., handball according to a natural or unnatural positioning of the player's arms). It could almost be said that
the rules of football have demonstrated Suber's Thesis:
any rule-based system that allows unrestricted self-modification eventually ends in paradox [Suber].
Given the ambiguity and uncertainty in the specification, interpretation, and application of the offside rule,
the effectiveness of SmartFlag offers an inversion of the
traditional approach, from rule specification and situation classification by the referees, to what we have
called algorithmic definition, whereby a situation is
defined as offside if the algorithm implementing the rule
outputs that the situation is offside. We then considered
the social and juridical implications of this approach, as
well as the implications of "machine learning maximalism" and marginalizing some data and (all) humans
from the decision-making process.
However, in keeping with the precautionary principle,
which implies taking social responsibility for the deployment of technology that might cause public harm, we
have attempted an analysis of the social implications of
decision-taking technology in the resolution of sporting
decisions; especially those decisions, like offside in soccer, which require a degree of interpretation by the referee (rather than a clear-cut binary decision, like in/out
in tennis).
In conclusion, we have found ourselves to be ambivalent with regard to the notion of "algorithmic definition" that could be provided by technology such as
SmartFlag. Algorithmic definition surely brings a sense
of consistency (if not transparency) to the application of
rules, would reduce the intense pressure on match officials at the highest levels, and might lead to fewer grievances manifesting themselves as abuse and violence.

IEEE TECHNOLOGY AND SOCIETY MAGAZINE

IEEE Technology and Society Magazine - December 2019

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - December 2019