Here is the "Deception" item.

Issues in Deliberate and Overt Deception

Sun May 31 20:34:34 1998
From: "John Mordigal" From Chuck LoPresti (d38716@gaviota.pnl.gov):

Ethical Guidelines: Section D, Item 5. Not everyone is familiar with why
one would want to practice "deception" at all; perhaps a clarification
using examples such as double-blind studies or placebo effect would be
appropriate here. [Charles LoPresti also commented on ethics regarding
statistical software. See that category for his remarks.]
------------------------------------------------------------------------
-----
Tue, 26 May 1998 09:43:04 +0000
Pat Nahas, following earlier correspondence, wrote:

I checked a dictionary and wonder if it might prove wise to include a
definition of deception in the Guidelines, with or without an example,
perhaps in a footnote. In my World Book Dictionary, listed as a synonym
for deceit, deception is described as follows: "Deception applies to the
act that gives a false or wrong idea, but does not always suggest a
dishonest purpose." As I think I mentioned originally, problems arise
when we have words or situations that can be interpreted differently by
different people. And I would venture to say that most people will not
consider the "does not always suggest a dishonest purpose" addendum to
the meaning. I just think we don't want to convey the wrong impression
or leave ourselves open to interpretation in this matter.

All comments received are referred to the Committee on Professional
Ethics; here is a preliminary response from Dr. Gardenier:

There is some disagreement even within the Committee as to whether
"double-blind studies or placebo effect" involve "deception." In
federally funded clinical trials governed by Institutional Review Boards
(IRB) and the Code of Federal Regulations, there must be prior informed
consent of patients to any such trials. Admittedly, this is not always
implemented perfectly. Also, there are clinical trials in the private
sector which are not subject to those rules. There may or may not be
equivalent rules in place. Finally, there are claims that clinical
practices, as opposed to clinical trials, contain far more deceptive and
risky experimentation with patients than do any clinical trials where
scientists are conscientiously attempting to obtain credible evidence of
treatment effects. This issue is far beyond the scope of a statistical
ethics committee. We encourage other, medically based bodies to examine
this problem and provide general ethical guidelines.

Pat Nahas pointed out that the World Book dictionary defines deception
as not necessarily involving dishonesty. Let me posit two examples of
common deceptive practice, which claim the moral high ground. You are
all invited to comment further on "ethical uses of deception, if any."

Statistical surveys of social issues, especially of sensitive matters
like family income, sexual behavior, underage drinking, illegal
activities, religious beliefs, and other matters cannot be accomplished
with any credibility except on conditions of the strictest
confidentiality. Conversely, in order for such studies to achieve
maximum scientific use, microdata tapes have to be made available to a
wide variety of researchers. As a first step in preparing such data for
release, all obvious identifiers such as name, address, telephone
number, social security number, etc. are stripped from the file. Still,
other factors in the data set may come together to enable identification
- especially such things as identification of a small geographic area, a
fairly precise time of an event such as a birth, an accident or a court
case, or unusual combinations of age, income, health factors, education,
or other attributes which are not in themselves obviously "identifiable
data." In general, the larger the number of records in a microdata file,
the greater the risk that one or more individuals in that file will be
indirectly identifiable, at least with fairly high probability.

Responsible survey organizations, including federal statistical
agencies, guard against such incidental violations of confidentiality.
Included within the body of measures available for such protection are
manipulations of the data such that some or all of the released records
contain false data. (Data may be swapped among records or whole
categories of data may be mathematically transformed in all the
records.) When such steps are taken, every effort is made to preserve
the statistical validity of the records overall and for important
component groups, such as by age, sex, race, and income level. Also, the
fact that the records have been manipulated to preserve confidentiality
is announced in the data documentation. After all, we do not want the
users to attempt identification and fail; rather, we want them to
understand that any attempt at individual identification would be not
only unethical, but also futile. This is "honest" in the sense that the
manipulation is announced before analysis proceeds; individual data
records, however, are undeniably deceptive.

Psychologists need to use broad categories of deceptive practices to
perform statistically valid experiments on how the human mind works.
Their ethics codes address not the deception in itself, but rather the
potential effects of the deception on the mental or physical health of
the subjects. One example: the same test is given to three equally
capable groups of schoolchildren. You tell one group, just before the
test, "You have been selected to take this test because you are
exceptionally well suited to it. (A lie.) You will most likely do very
well." Another group is told, "You are getting this test even though I
have tried to explain to them that you are nowhere near ready for this
type of test. (Another lie.) It seems unfair to give it to you now, but
that is what I have to do." The third group is simply given the test
with no biasing introduction. The null hypothesis is that the
test-giver's expressed expectations have no effect on student
performance. The results, of course, refute the null hypothesis. All
students are told the true nature of the test after the fact and learn
the real lesson is that self-confidence affects performance. To me, this
is an ethical experiment which is intentionally deceptive. Psychologists
can offer many similar examples. There have been, of course, deceptive
psychological studies for which the validity is inexcusable relative to
the potential or actual psychological harm to the subjects from the
experiment itself. The statistical ethicist's problem is to allow
justifiable deception without condoning harmful treatment of subjects.

The Committee will have to work on wording to accomplish that. Your
suggestions would be most welcome.
_______________________________________________________

Deception practiced covertly in the name of other ethical concerns, such
as confidentiality.

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

Dear Dr. Gardenier,

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 concerns section G.3. Although in most cases professional
discretion is proper, I can imagine that the G.3 requirement of
confidentiality might conflict with G.5 (the protection of
whistleblowers) and even G.4 (the reestablishment of normal professional
life). [His second question concerns unethical clients and is addressed
elsewhere.]

To go to some recent cases, when Henry Needleman (who first linked
elevated lead levels in children to learning deficits) was accused of
misconduct by members of the lead industry, the University of Pittsburgh
appointed a committee of faculty members to review his statistical
analyses. They eventually decided that his work was both honest and
professionally competent. Given that the case attracted wide media
attention, I think that Needleman's wish to make the full body of
evidence and testimony available to the public was an appropriate way to
recover his reputation. The imposition of a confidentiality rule in
high-profile cases would prevent the most honestly forthright response
that those who are incorrectly accused can make.

Although less statistical, the Baltimore/Imanishi-Kari case pointed out
another undesirable consequence of broad confidentiality rules. As you
probably know, there was widespread criticism of MIT's apparent use of
its internal confidentiality policy to bury Margot O'Toole's complaint.
This case is clearly complex, but it raises the issue that
confidentiality rules may sometimes be used as part of an institutional
whitewash of prominent scientists, or just as a means to stifle bad
press without consideration of the reputational costs to those caught in
the process.

More generally, I am reluctant to put forward a policy that prevents a
statistician from telling the truth. We are often called upon to
participate in tenure decisions or make recommendations about awards.
Although I certainly hope that all of us have the charity of spirit to
understand that people can make mistakes and learn from them, there is
still a strong duty on our parts to provide full information in such
deliberations. I would be uncomfortable if I felt that a confidentiality
rule were forcing me to collude in concealing unethical behavior.
------------------------------------------------------------------------
----------
Reply from Dr. Gardenier:

This illustrates a reason some philosophers argue that all documented
ethical codes or guidelines are futile. In practice, ethical issues
often become complex or even self-contradictory. For example, one is
often, but not always, relieved of ethical obligations when someone is
literally holding a gun to your head. Not every ethical situation is
worth dying for. If the "gun to the head" is figurative, rather than
literal, as in cases of getting your degree, keeping your job, obtaining
tenure, being allowed to conduct the year 2000 Census scientifically,
etc., there are generally important considerations of law and politics
involved, as well as ethics. One of the subtleties of ethics generally
is that institutions or powerful individuals can sometimes use ethical
arguments to promote unethical outcomes. Without commenting on the
specific cases noted above, no ethical guidelines can spell out exactly
how all ethical conflicts should be resolved. That must be left to the
maturity, judgment, and personal integrity of the individuals involved.

Possible suggestions include: confronting the officials with the ethical
conflict or advocating rules of procedure which lean heavily toward
protecting confidentiality while a case is under investigation, but lean
more heavily toward revealing the truth once the case has been resolved.
(This may be of little value when the cases take years to resolve, of
course.)

Further comments and suggestions on issues of deception will be most
welcome. You may post your comments here or e-mail them to Dr.
Gardenier at drgarden@erols.com.
>