Unethical Clients

>From - Sun May 31 20:30:28 1998
From: banks@cam.nist.gov (David Banks)

I want to congratulate your committee on its work in developing ethical
guidelines, as reported in the May 1998 issue of _Amstat News_. I
support nearly all the points you make, but have two small questions.
[The first concerned deceptive use of confidentiality restrictions and
may be seen under the heading, Deception.]

My second question concerns what obligations a statistician has when
they believe that a client is acting unethically. The 1989 draft
(especially article II.D.7) and commentary by Bross, Gehan, and
Mosteller, as well Deming's code, are somewhat at odds. Generally, it
seems that when a client intends dishonesty, the statistician should use
all diplomacy to persuade them to do right, and then withdraw from the
project if that effort fails. But the statistician is not allowed to
break the confidence of the client, especially to funding agencies or
university officers.

Clearly, one needs the client's trust in order to give the best
advice---almost every experiment has small failures of randomization,
protocol violations, and so forth, and a good consultant can often work
around these problems. But the client will not be frank if they fear
that any lapse will be reported to the university's ethics committee or
their grant officer. I've had cases where I urged my client not to
publish, or to include a clear statement about the violation of
analytical assumptions, and was ignored. Ought I to have done more than
withdraw?

We do not have the legal protections of lawyers and doctors on
confidentiality. I think that if we feel that our silence would entail
clear damage to others (e.g., in clinical trials) then we should betray
our client's trust. But absent such compelling circumstances, it is
probably improper for us to whistle-blow on our clients or employers.
But your group may have a very different view on this, and it would be
helpful to include more detail than the short statement made in section
B.7 (which almost appears to apply only to proprietary data or privacy
concerns).

As an ancillary point, when one withdraws from a project, what is the
appropriate way to deal with the results that have already been
obtained? Do these belong to the client, or should the statistician
attempt to prevent their publication? And if the statistician should
not block their publication (which may be part of a complex, multicenter
study that is mostly unimpeachable), should the statistician forbid the
client to acknowledge the statistician's work, or would that damage a
principle of accountability?

These are hard questions, and I'm sure there are many possible
viewpoints---the 1989 ASA Guidelines largely ducked these problems. But
please accept my thanks for the enormous body of important work that
your committee has achieved.

Sincerely,
David Banks
_____________________________________________________
From: Dr. Gardenier - Again, we have issues of law and politics mixed
with the ethics; still, we need to consider whether the Guidelines can
or should offer more help than is in the 1989 edition or the original
draft of the current revision. Comments? Suggestions?