Driver Trett Digest Issue 25 - Flipbook - Page 17
DIGEST | ISSUE 25
The italicised narrative you have just read, was not written
by me, but was produced by an AI chatbot created by San
Francisco based OpenAI, in response to a request posed
by me for it to “write me an article on the measured mile
analysis of productivity loss in construction”.
It took about 3 seconds get a response.
The ‘artificial’ opinion is not bad and offers a sensible sounding
synopsis of the measured mile analysis, and certainly passes
for media content, oven-ready for distribution on social
media.
Impressed, I asked the bot “is a measured mile approach
preferred over earned value in productivity loss analysis?”,
certain that a thorny opinion query on a preferential
analysis approach, that often may allude even the seasoned
professional, would surely trip the bot up.
The response, verbatim was:
A measured mile approach is a method used to determine the
productivity of a construction project by comparing the actual
output to the estimated output over a specific time period,
usually a week or a month. This approach is used to identify
the factors that are contributing to any productivity losses on
a project and to develop a plan to improve productivity.
Earned value management (EVM) is a project management
technique used to measure progress and performance on a
project. It involves measuring the value of the work that has
been completed and comparing it to the planned value of the
work that was expected to be completed at that point in the
project.
My answer (unsurprisingly you might say!) is no.
And it starts with looking behind the initial response. Whilst
both bots provided pretty impressive summaries about
productivity loss, they were just that – summaries. They
didn’t look at the practicalities and difficulties inherent in
demonstrating and measuring productivity loss, describing a
process rather than understanding it.
Indeed, the second ‘article’ produced by the AI did not
actually answer the question posed, namely, to indicate a
preference for a given approach. In my experience, those of
us who frequently provide such analysis will ordinarily have
a preference, and that preference is often governed by things
like the nature and quality of records available to conduct an
analysis, as well as other external factors.
Further, in my experience, a claims consultant or expert
witness will not require prompting to advocate the merits of a
favoured approach in a given set of circumstances. It may be
that to readily express a preference is a uniquely human
trait that would evade AI software, and by extension, bring
into question the usefulness of AI in expert opinion.
The responses generated by the bots indicate that the AI
software cannot (yet) produce an article that indicates a
practical comprehension of the subject; rather it provides
an illusion of comprehension. The result is a narrative which
might well be acceptable in less scientific publications, but
which is insufficient to ‘pass muster’ under any greater
scrutiny.
Conclusion, the bots may provide a reasonable introduction
to any subject, but it is far from being able to actually produce
an analysis. I think it’s safe to say that the bot is unlikely to
put an expert witness out of work, now or ever.
Both the measured mile approach and EVM can be useful
tools for analysing productivity losses on a construction
project. The choice of which method to use may depend on the
specific needs and goals of the project and the preferences of
the project team.
Again, this bot took about 3 seconds to respond, and again, it
was a pretty solid effort.
So, what does AI mean for construction
professionals and, in particular, those of us
who grind out a living in the construction
dispute space?
Are we about to be usurped by computers?!
17
