ANOMALOUS HUMAN-COMPUTER INTERACTION (AHCI): TOWARDS AN UNDERSTANDING OF WHAT CONSTITUTES AN ANOMALY (OR, HOW TO MAKE FRIENDS AND

• Document Type: CREST
• Collection: STARGATE
• Document Number (FOIA) /ESDN (CREST): CIA-RDP96-00792R000701020005-3
• Release Decision: RIFPUB
• Original Classification: U
• Document Page Count: 3
• Document Creation Date: November 4, 2016
• Document Release Date: May 17, 2000
• Sequence Number: 5
• Content Type: RP

Approved For Release 2009118/15: CIA-RD 1792M0b1O h20005-3
human
n
h
on p
seed informa
pro
side to this debate suggests that errors are caused by
cuts.
e
t
y
adaptive information-processing s
.,++r;r,,,tional field into that of human
sear
ment and decision making in the
d
g
ju
era have pointed out many weaknesse
ha
h
d
paper
esc- - Ps
h
Kahneman and A. Tversky (moo
Thus i
resent the more general processes t
cognitive errors.
certai
rocessing influences. The getnerallye
ti
Kahneman and Tversky sugg
are effecti
of events may b
objects or events,
availability
in perception, men
vidual's judgment of
Wends very much on th
strategy which can lead to
adju stment. Here, individ
ing with an initial value o
justed to account for new
which erroneous beliefs
confirming information.
logical research that
human judgment and
of errors in conclusi
settings. The rese
ent theory of huma
and present them
logical research on
parapsychologists
false-negative con
an
sions about the operation of p-
to stress that
quality of data on the occurrence of psi; and (3) mistaken
while there is some emphasis in parapsychology
g information--this is one way ,,
? of d s-
ibes three areas of
ognitive social psycho-
ear on the question o
ference and consequen
y on the examination
error and indeed, may not b
_P findings of relevance t
This re-
parapsychology
ii the vui-h'. ? .. ,
lusions about the occurrence of psi may be
istaken conelu-
help in eventually identifying m
hancinL the
in human cognition, oul Lhu?
been highlighted by D.
al Review, 1973, 237-251).
on-processing factors may rep-
st that people habitually use 1
o lead to errors. The first of
s
availability, where people's
t b
y
nenev of objects or the likelihood
to the accessibility of items
--;n
-
g
The
~1naLiOn .
---
sor similarity, where an indi-
s
~tivene
that two events are related de-
bilit
y
a
ee to which these events have features
The final information-processing
rrors is judgment by anchoring and
r tart-
o
conclusions that psi has occurred, psychological researc
error logically cuts both ways, and can aid in the identification of
false-positive and false-negative conclusions about the occurrence of
psi.
ANOMALOUS HUMAN-COMPUTER INTERACTION (AHCI):
TOWARS
(OR, AN
HOW TO MAKE FRIENDS AND INFLUENCE
COMPUTERS)
K. Morgan (Dept. of Psychology, Univ9Jty of Edinburgh,
7 George Square, Edinburgh EH8
This paper is an attempt to clarify in what manner a genuine
anomaly can be distinguished from an incident explicable by
physical means. It also tries to exemplify the various methods that
could be used to simulate an anomalous human-computer interaction
(AHCI). This paper does not dwell in any more than a superficial obs manner upon the psychology involvedi in manipulati go bservrrs
which would allow the described phy
out. That would demand a paper in its own right.
Part of the research being carried out at the Koestler Chair
and other institutions is the investigation of anomalous human-
computer interaction. As with any area of parapsychological re-
search there always exists the danger of the researcher mistaking
a normally explicable phenomenon as an anomaly. paper s
written to help people who are confronted by an unusual happening
to evaluate the situation
on a computer sibility of there beng normal methods offsimulating almost any pos
anom aly .
The various categories into which both simulated and genuine
anomalies could fall can be separated into the following:
(1) Human. The majority of so-called anomalies might be
found to be caused by the users' ignorance of their own computer
system or aspects of it. This, coupled with the human trait of
forcing unconnected events into meaningful patterns, might explain
many anomalies.
(2) Software Anomaly. The methods of achieving the simula-
tion of a softwar dware-based) anomaly can be broken down
into the following categories:
(a) Replacement of the target program. The target program
or process is exchanged for an amended version that contains the
extra "feature" that will become the "anomaly."
Approved For Release 2000/08/15: CIA-RDP96-00792R000701020005-3
 Approved For Release 2000tGOP15 : CIA-RDP964070RO0O 01020005-3
(b) Adjusting or amending the target program or process.
In this scenario the target process is amended at the same time or
very shortly before the "anomaly." This would of course demand
the knowledge of the specification of the software being used at the
target site.
(c) Breaches in computer /organizational security both prior
to and simultaneously with anomaly. These are often a necessary
prerequisite of any of the above software anomalies. The breach in
security could even be from a remote site, via a computer communica-
tion link that has access to the target system. The use of inter-
preted languages or online debugging (disassembling) tools makes
this reasonably easy.
(3) Hardware Anomaly. This section covers any physical ef-
fect that occurs to an item of computer machinery (but not to the
logic controlling it). These "effects" could often take the form
of either repairing or destroying an item of equipment. These items
could be anything from a personal electronic belonging (e.g., an
electronic watch) to a computer's storage device.
(a) Replacement. In this scenario an exact duplicate of the
target is prepared and exchanged for the target item when the op-
portunity arises. The duplicate has some extra "feature" which will
be used by the false anomalist to simulate the required anomaly.
(b) Adjustment/destruction--"live." To adjust or amend an
item of equipment is not as difficult as it might appear. Much com-
puting equipment is sensitive to one or many of the following en-
vironmental influences : weak magnetic fields, physical force (e.g.,
bending), exposure for long periods to strong sunlight, contact
with sharp objects, extreme humidity, temperatures outside the
tolerated range, contact with static electricity, and any substances
or object making contact with a recording or electrically conductive
surface.
(c) Breaches in computer /organizational security before or
simultaneously with anomaly. In contrast to the previous scenarios,
simulating hardware anomalies requires the physical presence of the
false anomalist or environmental influence in order to achieve the
anomaly.
(4) External to Computer System. Such things as electrical
mains fluctuations are a possible example of natural "disasters,"
and if the fluctuation coincided with some other meaningful event
the users might decide that an anomaly had occurred which had a
correlation with that meaningful event, thus starting local lore
about this false correlation.
Methods of Avoiding Computer-Based Fraudulent Anomalies
Examples of new technology that might help alleviate the above-
mentioned problems are
(1) WORMS (write once read many times) optical disks.
These (at present) are noneditable and are immune to "grubby
thumbs," magnetic fields, and static. They are therefore much
better potential psi-corruption targets than the currently favorite
floppy disk, especially if the target data on the disk are well en-
crypted and the disk uniquely identified.
(2) Optical fiber cables. This makes data transmission line
monitoring or adjustment much more complex.
(3) Automated technical advisers for computer-based security.
These can rapidly and thoroughly analyze a large and highly com-
plex system specification for security weaknesses. They are only
as good as the level of detail or accuracy in the specification and
the expertise of their user.
(4) Gypsy verification environments & (5) Program analysis
tools. Both of these methods could be useful in the analysis of a
piece of code that has been in an "anomaly." Again, they are sub-
ject to the same weaknesses discussed under the previous heading.
(6) Cryptographic methods. These can be highly effective
in preventing access to information, provided a sufficiently good
encryption method is selected.
(7) Shielding. Simple shielding of vdu screen emission can
eliminate the chance of a computer screen being reconstituted out-
side the system confines.
This paper has tried to portray the various scenarios that
could be misinterpreted as an "anomalous" human-computer inter-
action (AHCI). It also tried to show that there are conceptual pat-
terns that allow AHCI anomalies to be categorized, along with their
possible fraudulent explanations. It is hoped that armed with such
a method of categorization experimenters may be able better to re-
cord and evaluate the intriguing field of AHCI. In such evaluations
it may be more cost effective to create a means of detecting a fraud-
ulent anomaly rather than to proof a system against every possible
threat. A highly motivated false anomalist with large financial and
time resources might be able to create fraudulent anomalies, regard-
less of the tightest precautions. Experimenters might therefore
find it helpful to adopt an experimental condition where no one
Approved For Release 2000/08/15: CIA-RDP96-00792R000701020005-3
 Approved For Release 2000/08/15: CIA-RDP96-00792R000701020005-3
Papers
star "makes or breaks" the results. By using large "anonymous"
source groups the incentive for any one individual to create false
anomalies might be greatly reduced.
WHEN WILL WE BEGIN TO REDUCE ALPHA AND BETA ERRORS
IN STATISTICAL PSI EXPERIMENTS?
Ulrich Timm (Institut fur Grenzgebiete der Psychologie and
Psychohygiene, Eichhalde 12, 7800 Freiburg i.Br., West
Germany)
In many psi experi
made, after whose correct
appears. These are Type
That does not necessarily
real psi effects do not exis
correctly, are often so inef
instability, and inconsistenc
seldom lead to statistical sign
cal methods creates Type II er
tical selection errors are
'statistical significance dis-
er, that in these experiments
e usual methods, if utilized
ith regard to the rareness,
e. This inefficiency of statisti-
or beta errors. Therefore,
gnificances but also the reduction
ase of real significance.
First I give an overview of tho e alpha errors that I call
statistical selection errors. T ese sho , simply stated, the follow-
ing three qualities (Timm, 4, 1983, 19k-229):
(1) From a set of st istical result a single result is se-
lected and eval ted by some si ficance test.
(2) The selection not performed radomly but according
to a criterion hat is related to th level of the single
result in that it directly or indirectly favors positive
(3) Despite this success-dependent selection, the significance
test is carri~d out and interpreted in the usual manner
without any correction.
Approved For Release 2000/08/15: CIA-RDP96-00792R000701020005-3