Expert Witness Journal Issue 67 June/July 2026 - Flipbook - Page 99
particular individual. Perhaps death rates have been
unusually high, or perhaps not. Perhaps a medical
practitioner is unorthodox or has upset some of
their colleagues.
To produce reliable evidence, science uses rigorous
controls, precise de昀椀nitions, robust statistical
analysis, and adopts the hierarchy of evidence
to prioritise the highest quality data over mere
‘opinion’. Of course, real world investigations
would be untenable if scienti昀椀c gold standards had
to be met in all circumstances. There is a problem,
however, when ‘science’, in the form of expert
‘opinion’ evidence, is relied upon when little more
than credentialled bias.
Once an investigation has begun, it could end in a
criminal prosecution even in the absence of reliable
evidence. There is an all too human tendency to blame
people for bad outcomes. People generally have a
poor intuitive understanding of statistics. Group
think creates a snowballing e昀昀ect: as investigations
gain momentum, biases shape understanding and it
is ever less likely that dissenting voices will change
its course.
In most cases, for example, prosecution experts
should be subject to strict ‘blinding’ controls;
their evidence independently assessed rather than
adopting methods that encourage ‘group think’;
professional statisticians should be instructed much
more frequently; experts should be held to their
duties to act independently and impartially, to
make clear when issues fall outside their expertise,
to identify anything that might undermine their
opinion and any reasonable range of opinion –
failing which they should not be excused in order to
‘keep the show on the road’: their evidence should
be rejected.
A blind eye
Once a case comes to court, the usual approach is
to pretend that cognitive bias does not exist or is
inconsequential: These are experienced experts
who understand their duties to the court. It is for
the jury to assess the evidence: they can choose to
give less weight to the evidence. Or a simple 昀椀nding
that there is no evidence of bias, especially if it might
derail proceedings to 昀椀nd otherwise.
Courts should understand that when there is dispute
between experts, it is best resolved by evidencebased methods, rather than a mere exchange of
expert ‘opinions’. If high quality data is available,
opinions should be based on a structured review of
high quality data, not ‘cherry picking’ lower quality
data (or even misrepresenting data) to support their
case.
The misunderstandings here are that bias is
something exceptional rather than part of the
human condition; and that bias can be identi昀椀ed and
managed by ordinary common sense. If it could, we
wouldn’t have been lobotomising patients late into
the twentieth century. Or handing over billions of
our savings to (with hindsight) obvious fraudsters.
Or regularly locking up people for crimes they did
not commit.
Of course, the justice system is populated by
lawyers (and former lawyers) and not scientists and
statisticians. There will be incomprehension and
resistance to such proposals. We do, however, have
an adversarial system. It will be the role of some
lawyers, and the experts they instruct, to persuade
other lawyers (and former lawyers), or the jury
directly, that opinion is not evidence: that what
matters is the quality of the evidence underlying
that opinion. That any expert who does not follow
evidence-based methods, who does not comply with
their expert duties, should be rejected.1
The UK’s judicial system is not without safeguards.
We have an adversarial system, so that di昀昀erent
perspectives are heard and contested. We have
independent arbiters of fact. We have burdens of
proof. We have certain rules of evidence, such as the
exclusion of evidence which is more prejudicial than
probative.
To some degree, judges might reasonably say: It’s
not perfect, but this is the system. My role is to help
it function, guarding it from those who would attack
its foundations. We must, after all, live in the real
world.
Reference
1
For further discussion of this topic, read the excellent report
by the Royal Statistical Society, ‘Healthcare serial killer or
coincidence? Statistical issues in investigation of suspected
medical misconduct’. See also the Forensic Science Regulator
Guidance on ‘Cognitive Bias E昀昀ects’.
What is wrong with that?
There have been various reforms to help guard
against cognitive bias since, for example, the
Birmingham Six were wrongly convicted in the mid1970s, not least the introduction of PACE 1984 to
protect against false confessions. The unscienti昀椀c
approach to expert evidence however leaves a major
risk of miscarriages of justice.
EXPERT WITNESS JOURNAL
97
JUNE 2026
