Page 179 - Cyber Terrorism and Extremism as Threat to Critical Infrastructure Protection

GRAEME BALLARD: IF THE FACE FITS: IS IT POSSIBLE FOR ARTIFICIAL INTELLIGENCE TO ACCURATELY PREDICT THREATS TO
PROTECT CRITICAL INFRASTRUCTURE?

best results from both quantitative and qualitative methods in a ML tool, as long as certain
well-established principles are maintained and applied.

4.2 (Much) Wider Contexts

Section 1.1 discussed the relevance of energy production and consumption to the formation and
security of the modern nation state, and the inherent conflict therein. It is inevitable that these
conflicts and power struggles must be approached and reflected in the analysis, if ML is to be
used effectively as a revolutionary, predictive tool. This means we must not only consider the
models within the scientific literature – in this case, psychology – but also explore the links be-
tween the disciplines themselves: biology, psychology, sociology, and the very history of human
beings, as per the Figurational approach (Dunning, 1999). The mechanisms used by the different
groups in society for obtaining power, as well as the power conflict itself, are subject to constant
change, which affect the outcome of power struggles in society and are relevant in the analysis
of threat. Power, in this sense, becomes a social process in and of itself (Dunning, 1999) which
can be treated very much like a process within psychological theory, such as arousal.

The implications of this, however, are more complicated than it might at first seem, because
it means we must also account for the way we view and understand the world scientifically
(Dunning, 1999). It can be argued that scientific facts, as we commonly know and understand
them, are actually part of the social process of power management. The hypothetico-deductive
reasoning behind science and the scientific facts that emerge, can themselves become viewed
as a power-management tool (whether deliberately, accidentally, or incidentally). Such power
management might not be a deliberate act, but an unintended consequence of the process of
functional unity and falsification required of the quantitative method.

In practical terms, this is one of the explanations of why threat, in and of itself, might be
defined differently between countries or regions (e.g. differences in the way health is defined).
Furthermore, within regions and countries, what is defined as a threat will change over time
(e.g. what constitutes a terrorist threat), and these same societies might temporarily modify or
suspend what is viewed as a threat in certain situations (e.g. policy and behaviour in the face
of the Covid-19 threat).

Once this idea has been accepted, then it is possible to analyse and critique certain values
and beliefs that quietly permeate the whole of society, usually without much thought. For
example, there are widely held beliefs that women do not commit sex-offences (Gillespie
et al., 2015) and that athletes using drugs are cheats and morally inferior (Van Raalte et al.,
1993). These assumptions might not seem unreasonable until it is considered how, when
taken as a whole, they genuinely effect the integrity of research, the legal system and the
rehabilitation protocols relating to female (and, possibly, male) sex offenders (Williams et
6
al., 2019), or the integrity and success of drug-testing programmes and the potential negative
health consequences, related to drug testing itself, in the world of sport (Ballard, 1999). Bad
data input always equals bad data output – whether using legacy or ML analytical tools.

Ensuring that only good data is used and maintained in a ML tool will be far more important
than it has ever been before, due to the leap in efficiency inherent in ML. This efficiency

6 Although I’m specifically speaking about the sports context here, the definition of success in any drug testing
programme, whether workplace or criminal justice system, is highly debatable.

179