This paper identifies yet another field of research, the discipline of huma
n computer interaction, where the concept of self-similar fluctuations can
play a vital role. A concept of interaction between computation and cogniti
on is developed that is friendly toward the cognitive process. It is argued
that friendly interactions must have a memory and be antipersistent. To ca
st this in a mathematical form, fluctuations in the interactions recorded o
ver a period of time are studied, and it is shown that these fluctuations m
ust necessarily be self-similar with the value of the self-similarity param
eter confined to the interval (0, 1/2), for the interaction to be friendly.
A statistical measure of complexity, of the interaction process, is also f
ormulated as a function of the self-similarity parameter. Finally the quest
ion is raised as how to build a friendly software and a possible evolutiona
ry process through which friendly softwares may emerge is indicated. (C) 20
01 American Institute of Physics.