, , , , ,

Previous blog posts in this series:

Part 1, Restating the problem

Part 2, The I in you

Part 3, Meta-modelling


So far I have laid out the problems are that arise when we talk about our (future) interactions with distributed, not too intelligent, artificial agents (part 1), and I also investigated the underlying reasons for this issue (part 2 and 3). In this blog post, I am compiling my ideas about how we could overcome our shortcomings by thinking about this topic rationally but in different ways than discussed so far. In a forthcoming blog post I will address approaches that rely on alternative to rational thinking.

Note well: In order to make this blog post less wordy I refer to distributed, not too intelligent, artificial agents -as usual- as “artificial idiots.”

Good explanations

So what would a good way of talking about our interactions with “artificial idiots” look like? The short answer is: a way that makes us understand our relationship, our interdependencies better. But what do we mean by “understanding?” Let me outline what I think understanding “artificial idiots” should encompass.

What we are looking for is an explanatory method that comes with at least three properties. One property is the utility of prediction: given information about how artificial agents are constructed, how they behave, how they interact with “the world,” how we behave, we should be able to predict how our interactions with “artificial idiots” would look like, and how the mutual interaction would transform both us and the “artificial idiots.” Second, the method should provide guidance on how to decide problems we are faced with. What capabilities should “artificial idiots” be bequeathed with? Who should own “artificial idiots?” Who should control their capabilities? What to do with the “artificial idiots” once their time of use has expired? One could of course answer that “artificial idiots” are developed for specific purposes (for instance, atmospheric monitoring in an urban area), so most of these questions are already answered by virtue of what problems the “artificial idiots” are designed for. However, such an answer misses the big pictures. Important and thorny questions will arise at a next higher level of generalisation, i.e. at the level of interaction between many (groups of) persons and many different types of “artificial idiots” or even entire ecosystems of “artificial idiots.” Especially when talking about learning artificial agents, there is a non-negligible potential that they eventually will transcend their initial task. Are they then allowed to change their task for a new one? What degrees of unintended autonomy are we willing to grant the “artificial idiots?” These are but a few of many questions one would face in such a situation.

Third, but not least, such an explanatory method should be “developable,” by which I mean that the method should be amenable to new insights, new findings, new facts. For instance, when we learn more about how immersion into a world sated with artificial agents changes our neurological wiring, or how we interact with each other (and the agents), this information needs to be fed back into the method and update it, calibrate it. Let’s not entertain any illusions: any method we can come up with at the current state will by necessity be rather a sketch than a developed theory. This is owed to the fact that the deployment of “artificial idiots” has only just begun and that there has not been much research into the emergent societal patterns of behaviour, the ecosystems that are triggered by connecting an increasing number of distributed sensors, actuators and, yes, “artificial idiots.”

So what explanatory avenues can be identified already today? Well, there is science, and there is philosophy.

Wake up, science!

In my previous blog post I was rather dismissive about science. Not that I believe that science cannot, in principle, provide us with methods for understanding our interactions with “artificial idiots.” Rather, it has not done so yet, and existing approaches (subject-object disjunction, sociology) cannot readily be extended to cover our interactions with “artificial idiots.” This lack of engagement is not because the wider technical community is not interested in this topic at all; for instance, the emergence of humanoid robots has been addressed early on [Brooks 2002]. I had a recent look at the pertinent literature, but besides narrowly focused studies (see, for instance, [Chen 2012]), I have not found anything that can help us with the issue at hand. In case you are aware of a publication I missed please let me know.

This does not imply though, that science would have to start from scratch, since seemingly unrelated topics can turn out to actually be central “artificial-idiots” topics, at least when taking a second look. One example is the role of group selection in Darwinian evolution and how competition, but also cooperation, between groups can contribute to a change in the survivability of individual organisms. While most of the research focuses on competition and cooperation within the same species (see, for instance, [Wilson 2007]), symbiosis between species has also been addressed (see, for instance, [Frank 1994] and [Doncaster 2013]). Albeit “artificial idiots” usually are not subject to a Darwinian evolutionary process, some of the models developed in biology might still turn out to be helpful in understanding our interactions with non-Darwinian “artificial idiots.” Also, as pointed out in the first part of this series, it is quite conceivable that some artificial agents will actually evolve through a process akin natural selection, which will make the biological models on inter-species interaction even more useful.

Let us talk about philosophy

Turning our attention to philosophy, the picture is actually quite different. Not that many of the philosophical efforts I am aware of already deliver answers to the questions raised in this series, but the questions are at least partially addressed in different fields within philosophy. These efforts can roughly be divided into two camps: “re-charting the map” and “inclusiveness.” Let me address them one after the other.

Re-charting the map:” new materialism

What I mean by “re-charting” the map is a larger undertaking in philosophy that wants to achieve a modern metaphysics free of Descartes’ core concept of dualism [Robinson 2012]. Instead of putting human beings centre stage, the philosophers in questions want to understand the world from the perspective of a material being (us) that is embedded in a material world.

One of the antecedents of new materialism was phenomenology [Smith 2013], which is our attempt to understand the world solely based on our sensations. While this still seems to put human beings centre stage, it removes the concept of human beings as a divine and superior “otherness,” as not “being in the world.” Instead, humans are part of an increasingly flat ontology. Relevant investigations into phenomenology were conducted by Merlau-Ponty, who focused on humans as material beings that are part of a flat ontology [Coole 2010a, p. 96]. This direction was adopted by new materialism, which emerged around the turn of the century. For a thorough discussion of new materialism and how it compares to post-modernism see [Coole 2010b].

2014-08-19 - phenomenology 2

In new materialism, humans are just one representation of a material “assemblage,” and categorically no value difference is made between humans, other animated beings, and non-animate objects. All of them are viewed through concepts such as their capacity to change [Coole 2013] and to interact with other objects [Kennedy 2013]. To my knowledge, “new materialists” have not addressed the topic of “artificial idiots” yet, but the methodologies they develop should definitely be applicable.

However, while new materialism moves the focus from humans to “material” and the connections, influences, and the dynamics that exist and emerge between material assemblages and within them, the philosophical turn of new materialism has so far been of a rather descriptive nature. It fails thus short in the guidance category I would expect from a good method (see above). Also, due to its descriptive focus, new materialism has so far focused on charting the material “landscape” and has, to my knowledge, not tried to develop predictive methodologies. In that sense, new materialism is so far mostly a “forensic” endeavour.

Re-charting the map:” information-centric metaphysics

While the new-materialism turn is mostly taking place in continental philosophy, there are also re-charting efforts going on that have their origin in analytical philosophy. One of the active thinkers in this field is Luciano Floridi, who has published extensively on the philosophy of information. Besides getting a better “grip” on what information is, and what foundational role it plays in today’s societies (and future societies), he has also embarked on an ethics of information [Floridi 2005]. His approach to this topic somewhat stands the world on its head. Instead of grounding his arguments in matter, as is done in new materialism, he builds up an entire metaphysics with information as its foundation. In this world view, organisms are imbued with value because they contain information [Floridi 2010, chapter 8]. In this different version of a flat ontology, the difference between “artificial idiots” and humans is also only gradual.

It is hard to fathom for me, right now, how other important aspects of a cognitive framework, e.g. the epistemology of “artificial idiots,” can be deduced solely from an assumption of the equality of information. To be fair, this is not what Floridi attempts, rather he wants to adapt and morph anthropomorphic concepts, definitions and views, so that they apply to “information content” irrespective of its embodiment and origin. However, it is maybe wrong to categorise this as a “re-charting” effort; maybe it rather belongs to what I characterised as “inclusiveness.” So let’s go there.


Inclusiveness is a label that I attribute to philosophical efforts that expand/modify existing non-metaphysical concepts so that they also encompass artificial agents (and other forms of matter). Such kind of contemplations are actually quite frequent, since we are increasingly faced with emerging questions, especially in the realm of “artificial ethics,” for which we do not have evaluative concepts nor ready solutions yet. What I am referring to here are, for instance, ethical dilemmas stemming from our tendency to automate more and more of our daily chores. Not only are we faced with questions of guilt attribution if something goes wrong (just wait for the first time a “driverless” car is involved in an accident) but also with questions of what ethics we should encode into the artificial agents we task with our daily chores. Just imaging the following situation, which is a variant of the Trolley problem [Wikipedia 2014]: a driver-less car is faced with a no-win situation, in which it either can plough through a group of pupils who suddenly appeared on a cross walk or swerve to the side and kill a commuter waiting at a bus sign. How are the algorithms supposed to resolve this situation, and who is entitled to make that decision? But the epistemic problem actually starts even one step earlier. The root question is whether artificial agents can commit moral acts and how we are supposed to react when the commit morally wrong acts (for instance, killing a person).

Rather than trying to solve this problem one time for all, many philosophers take a step-by-step approach [Allen 2000; Floridi 2004]. For instance, Floridi and Stalker decided to toggle what constitutes a morally wrong act and “only” widened the group of actors that we deem capable of committing moral acts. In their analysis, they come to the conclusion that any agent is able to commit moral acts, but that only a subgroup of agents can be held responsible for these very acts [Floridi 2004].

I agree that none of the above efforts provides the methods I have pleaded for at the beginning of this blog post, but all of them are at least tangential, and during the course of their development and maturation one can reassess them according to the quality markers of a good method, i.e. a method that

  • lends us the power to predict,
  • lends us the power to decide, and
  • which is “developable.”

After all, copying, rearranging, and combining existing ideas is what makes us progress as a society [Ferguson 2010-2012].

Next up …

Approaches that do not rely on rational thinking.


Thanks are due to Miranda Bruce at the Australian National University for directing my attention to new materialism [Bruce 2014], and for discussing this topic in depth, and for sharing her ample knowledge with me.


[Allen 2000] Colin Allen, Gary Varner, and Jason Zinser. “Prolegomena to any future artificial moral agent.” Journal of Experimental & Theoretical Artificial Intelligence, Vol. 12, No. 3, pp. 251-261, 2000.

[Brooks 2002] Rodney Allen Brooks. Flesh and machines: How robots will change us. New York: Pantheon Books, 2002.

[Bruce 2014] Miranda Bruce, “The Matter with Matter: New Materialist theory and the Internet of Things.” Conference on the Philosophy of the Internet of Things, York, U.K. Available: https://www.youtube.com/watch?v=zzXglBnjCAw&list=UU_TO7PeOZ1MWRqzN_xzOwmg (accessed: 2014-08-27), 2014.

[Chen 2012] Yen-Kuang Chen. “Challenges and opportunities of internet of things.” Design Automation Conference (ASP-DAC), 2012 17th Asia and South Pacific. IEEE, 2012.

[Coole 2010a] Diana Coole. “The Inertia of Matter and the Generativity of Flesh.” In: Diana Coole and Samantha Frost (Eds.), New Materialisms. Duke University Press, Durham & London, 2010.

[Coole 2010b] Diana Coole and Samantha Frost. “Introducing the New Materialisms.” In: Diana Coole and Samantha Frost (Eds.), New Materialisms. Duke University Press, Durham & London, 2010.

[Coole 2013] Diana Coole. “Agentic Capacities and Capacious Historical Materialism: Thing with New Materialisms in the Political Sciences.” Millennium Journal of International Studies, Vol. 41, No. 3, pp. 451-469, 2013.

[Doncaster 2013] C. P. Doncaster, A. Jackson, and R. A. Watson. “Manipulated into giving: when parasitism drives apparent or incidental altruism.” Proc R Soc B 280: 20130108. Available: http://dx.doi.org/10.1098/rspb.2013.0108, 2013.

[Ferguson 2010-2012] Kirby Ferguson. Everything is a Remix. Available: http://everythingisaremix.info/watch-the-series/ (accessed: 2014-12-04), 2010-2012.

[Floridi 2004] Luciano Floridi and Jeff W. Sanders. “On the morality of artificial agents.” Minds and machines, Vol. 14, No. 3, pp. 349-379, 2004.

[Floridi 2005] Luciano Floridi. “Information ethics, its nature and scope.” ACM SIGCAS Computers and Society, Vol. 35, No. 2, 2005.

[Floridi 2010] Luciano Floridi. Information: A very short introduction. Oxford University Press, 2010.

[Frank 1994] Steven A. Frank. “Genetics of Mutualism: The Evolution of Altruism between Species.” J. theor. Biol. 170, pp. 393-400, 1994.

[Kennedy 2013] Rosanne Kennedy, Jonathon Zapsnik, Hannah Mc Cann, and Miranda Bruce. “All Those Little Machines: Assemblage as Transformative Theory.” Australian Humanities Review, Vol. 55, pp. 45-66, 2013.

[Robinson 2012] Howard Robinson, “Dualism”, The Stanford Encyclopedia of Philosophy, Edward N. Zalta (ed.), available: http://plato.stanford.edu/archives/win2012/entries/dualism/, 2012.

[Smith 2013] David Woodruff Smith. “Phenomenology.” The Stanford Encyclopedia of Philosophy (Winter 2013 Edition), Edward N. Zalta (ed.). Avialable: http://plato.stanford.edu/archives/win2013/entries/phenomenology/ (accessed: 2014-08-28), 2013.

[Wikipedia 2014] Wikipedia contributors, “Trolley problem,” Wikipedia, The Free Encyclopedia, available: http://en.wikipedia.org/w/index.php?title=Trolley_problem&oldid=622736501 (accessed: 2014-08-28), 2014.

[Wilson 2007] David Sloan Wilson and Edward O. Wilson. “Rethinking the theoretical foundation of sociobiology.” Quarterly Review of Biology, 82:4, pp. 327-348, 2007.


If you like this blog post please click the below “like” button. If you want to stay up to date with my island letters please use Worpress’s “follow” function or the option provided in the panel to the right.

Also, please share your thoughts in the comment section; letters can and should be sent both ways!


Update 2014-12-30: Added a piece of my own art to light up this blog post.