Page 1 of 3
1
Autonomous causality and the under-appreciated
significance of stigmergy
Russ Abbott
Dept. of Computer Science, California State University, Los Angeles
Russ.Abbott@gmail.com
Abstract. How can symbols act as causes—as they do when for example, a
traffic light “causes” (under an interventionist interpretation) traffic in one
direction to stop and in the other to go? Unlike traditional causation in which
cause determines effect, the effect of a symbolic cause is determined by the
entity that responds to the symbol. I call this autonomous causality.
Stigmergy involves one entity leaving a symbol-like “trace” in the
environment for another entity (or itself) to encounter later. The party that left
the trace cannot control the response. Stigmergy involves autonomous causality.
Because computers determine the meaning of programs, they are
autonomous observers of the (software) “traces” left by programmers. A
programmer cannot force a computer do anything: computers do not do
“exactly what one tells them to do.” No matter how declarative a programming
language, a programmer must focus on the computer’s interpretation of symbols
and its consequent action rather than on intended results., Like virtually all
forms of communication, programming is stigmergic.
The humble programmer can do little more than put symbolic expressions in
the path of an oncoming computer and hope that the computer’s responses will
realize her goals.
1 Autonomous causation
Does it make sense to say that software, i.e., symbols, cause a computer, i.e., a
physical object, to act in a particular way? When put that way, probably not. This
paper explores an alternative perspective on symbolic causation.
Physical causality, Dowe (2000), involves the transfer of a quantity, such as
momentum, from cause to effect: e.g., one billiard ball hitting another. But:
Both (a) a court issuing an execution order and (b) the captain of a firing squad
commanding “Shoot!” cause, (Pearl, 2009, p202.), the death of the prisoner.
A traffic light changing color causes cars to start/stop.
Only vanishingly small amounts of any physical quantity are transferred: photons
striking retinas. The symbolic aspect of these events serve as causes. Yet,
1. Symbols are abstract and hence causally inefficacious. (Rosen, 2014)
2. Symbols have no intrinsic properties. A symbol’s only property is that it can be
distinguished from other symbols.
Page 2 of 3
2
A symbol produces an effect only when an agent capable of acting responds to it. I
call this autonomous causation. Responding agents determine the effects of symbolic
causes.
Laplace (1814) famously wrote, “we can regard the present state of the universe as
the effect of its past and the cause of its future.” Laplace said nothing about symbolic
causes. Instead of the laws of physics pushing the world around, autonomous agents
may “choose” how to respond to symbolic causes. (This resembles Dennett’s version
of free will—most recently, Dennett 2020).
An agent’s response depends not only on the symbol itself but also on the agent’s
history: how it was programmed, i.e., what it was “taught,” how the program may
have been modified or may have modified itself. Without information about an
agent’s history, its response to a symbol may be effectively undiscoverable—except
by giving it the symbol and watching what it does.
Even full knowledge of an agent’s history may not provide an effective way to
determine its response. Although the design and history of AlphaGo (Silver et. al.
2016), the computer program that beat the Go world champion, are known, the only
feasible way to determine its response to a board position is to give it that position
and observe its move.
When using a word processor one has the sense that striking, say, the “e” key
“puts” an “e” character into the text. Although that’s more-or-less how a manual
typewriter works, that’s not what happens with a computer. Striking a key transmits a
message to the computer’s software, which responds with some action, such as
inserting a character into a text. Software developers create the illusion that users
have direct control over the action of a computer. (See also Woodward, 2007 on
causation as remote control.)
2 Stigmergy
Grassé (1959) defined stigmergy as “traces” insects leave in an environment that
stimulate actions by other insects. Examples include termite nest building and ant
foraging. More generally, stigmergy can facilitate group coordination as in swarm
intelligence and ant colony optimization. (See Yang et. al., 2016)
Citing Wikipedia, Heylighen (2015) finds stigmergy to be a general coordination
mechanism. Readers are stimulated to improve and expand the writings of previous
contributors. Heylighen also credits stigmergy with underlying the “invisible hand” of
auction markets. Transactions affect bid and asked prices, which affect further
transactions.
I would categorize as stigmergic any interaction that includes a symbolic
component. Email messages and text messages are traces left for others; notes to
oneself are stigmergic communication with oneself. (See Clark and Chalmers, 1998.).
Even direct verbal communication is stigmergic. Sound waves are the traces.
Vaccination offers a timely example. A vaccine does not prevent disease: in
response to a injection, the recipient’s immune system generates antibodies, which do
the work.
Page 3 of 3
3
The originator of a communication does not control the response of the recipient.
(Abbott, 2018) In the famous cobras-in-Delhi example, the government established a
bounty for dead cobras. Citizens responded with snake farms.
Finally, one cannot command a computer do anything. At best, one can arrange for
a computer to encounter cleverly designed symbolic expressions and hope that its
interpretations and consequent actions will realize one’s goals.
References
1. Abbott, R. (2018). Meaning, autonomy, symbolic causality, and free
will. Review of General Psychology, 22(1), 85-94.
2. Clark, A. and D. Chalmers (1998). "The extended mind". Analysis. 58 (1):
7–19.
3. Dennett, D. C. (2020). Herding Cats and Free Will Inflation. URL:
https://ase.tufts.edu/cogstud/dennett/papers/Dennett_Romanell_C2020.pdf.
4. Dowe, Phil (2000). Physical Causation. Cambridge University Press.
5. Laplace, Pierre Simon (1814). A Philosophical Essay on Probabilities.
translated by Truscott, F. W. and F. L. Emory, Dover Publications (1951)
6. Grassé, P. P. (1959). La reconstruction du nid et les coordinations inter-
individuelles chez Bellicositermes Natalensis et Cubitermes sp. La théorie de la
stigmergie: essai d'interprétation du comportement des termites constructeurs.
Insectes Soc. 6, 41-81.
7. Heylighen, F. (2016). Stigmergy as a universal coordination mechanism I:
Definition and components. Cognitive Systems Research, 38, 4-13.
8. Pearl, J. (2009). Causality. Cambridge university press.
9. Rosen, Gideon, Abstract Objects, The Stanford Encyclopedia of
Philosophy (Spring 2017 Edition), Edward N. Zalta (ed.).
10. Silver, David, et. al (2106). Mastering the game of Go with deep neural
networks and tree search. Nature 529, 484–489, (28 January 2016),
doi:10.1038/nature16961
11. Woodward, J. (2007). “Causation with a Human Face” in H. Price and R.
Corry eds. (2007), Causation, Physics and the Constitution of Reality: Russell's
Republic Revisited Oxford: Oxford University Press.
12. Yang, X. S., S. Deb, S. Fong, X. He, & Y. Zhao (2016). Swarm Intelligence:
Today and Tomorrow. In Soft Computing & Machine Intelligence (ISCMI), 2016
3rd International Conference on (pp. 219-223). IEEE.