Previous blog posts in this series:
Part 1, Restating the problem
Part 2, The I in you
In my blog post “Restating the problem,” I re-introduced the topic of “artificial idiots” and my believe that our current approach to talking about the dynamics between us and “artificial idiots” is lacking and sometimes even defective. Furthermore, the importance “artificial idiots” already play in our daily life, and that they will play in the near future justifies a deeper investigating into why it is that our narrative about “artificial idiots” fails and how we might rectify this situation.
In the next blog post “The I in you,” I elucidated how we tell stories, why we do it the way we do, and what shortcomings our traditional narration has when it comes to artificial intelligence that is simple but pervasive.
In this blog post I am addressing why even a scientific approach toward our interactions with “artificial idiots” is burdened with shortcomings.
The problem – as I see it
The emergence of an augmented reality based on interconnected intelligent agents will change society. If we cannot narrate and analyse this change, we can of not portray what happens to us. Such a blind spot is consequential, since if we cannot formulate our observations, and we cannot formulate real solutions for emerging problems either. We would, in other words, be subject to -fundamental- societal changes without the capacity of conscious, rational, directed reactions to these changes. Our reactions would thus, by necessity, be inadequate.
But shouldn’t science be capable of overcoming this limitation? After all, science does not rely on, for instance, human agency and character as narrative anchors (see the previous post, “The I in you”). Rather, it offers a language that is made for describing non-animate and thus also artificial phenomena. This is right, but the problem with “artificial idiots” is that we need to understand our interaction with artificial intelligence, and in this case we are both subject (scientific interrogator) and object (actors in the augmented world). In this situation one cannot longer readily distinguish between “artificial idiots” and humans. In this case the traditional subject/object disjunction used by science no longer describes reality. Let me unpack this argument and let me present arguments in support. Be patient though, the unpacking is rather involved
As pointed out before, the ever-increasing abundance of interconnected intelligent agents will change the world and especially society. How do we usually analyse, portray, and narrate societal changes? In its simplest form, societal changes stem from interactions with “the other.” At the foundation of this very statement is the claim that an interaction can be broken down into a relation between two entities, a subject and an object. This split is usually called dualism [Robinson 2012]. Hereafter, I am referring to the concept of breaking down entities into subjects and objects as a subject-object disjunction. The operational relation between subject and object is henceforth referred to as pairing.
Roughly, all social interactions can be grouped into two major categories in terms of how the subject-object disjunction is expressed. First, there is our interaction with the non-computational, inanimate world (rock, apple, lever, Swiss watch …). Second, there is our interaction with the animate world, which can be broken down into two subsets. Category 2a addresses our interactions with non-human life forms (animal, plant, bacterium …) and category 2b addresses, unsurprisingly, our interactions with other human beings.
Yes, I have left out computers. I will address them when I finally address our interactions with interconnected intelligent agents.
So, how do these categories relate to the afore introduced subject-object disjunction and subject-object pairing?
Category 1 can be portrayed as interactions that adhere to a simple subject-object pairing (see the below Figure).
Panel (a) displays the simplest subject-object pairing possible; it is single-sided, no feedback occurs: A person plucks an apple (from a tree). We can readily distinguish subject (person) and object (apple). They are not the same and one operates on the other. Panel (b) displays the next simplest subject-object disjunction, which comes with a double-sided pairing. A person steers a bicycle and the bicycle carries the person. Here, the borders between subject and object are a bit obscured, but the two can still be separated. After all, the operations a person can carry out are not limited to steering bicycles, so the person as an entity has an existence beyond this particular subject-object pairing.
So what about category 2? Let me address the more complicated but conceptionally easier subcategory, which is subcategory 2b: interactions between people. One is tempted to apply the subject-object disjunction to these interactions too (you, the reader, are not I, the writer), but this disjunction fails to adequately portray and explain the wealth and strength social interactions. Let my explain this by aid of the next Figure.
Panel (c) portrays the double-sided interactions within a nuclear family (mother, father, single child) and the outside world (here: a relative). The nuclear family forms a unit: interactions within this system are more frequent and stronger than with persons outside this system. We know who belongs to the system (composition), how the members relate to each other (structure), how the family works (mechanism), and what lies outside the family (environment). Notice that the relative interacts with single persons at a time (child, mother, father), but the bounds between the members of the family are stronger than with the “outsider”, the relative. Therefore, in a sense, the relative interacts with the family as a whole when interacting with individual family members. The family is an epiphenomenon. This form of unit, i.e. the nuclear family, fulfils all the characteristics of a system [Bunge 2004]. Next to the above discussed subject-object disjunction and pairing, system theory is the second fundamental concept that is used for describing interactions in a scientific context. The family is an emergent unit, namely a system, and its members can not longer be treated as independent entities. This is already indicated by their labels. I am not your child, I am the child of my parents. This type of strong bond and interaction between humans is what gives rise to the emergent entity of group, which is a the focus of sociological studies..
So far, I have identified two types of interaction patterns: those that can be categorised by aid of a subject-object disjunction, and those where the units form such strong bonds that they cannot longer be understood and treated as separate entities, i.e. as individuals. However, even in the latter scenario not all interactions are equally strong and frequent, so that subsets of individuals can be identified and described by the aid of system theory, and pairing can then be applied to entire systems [Bunge 2004]. For examples, “family Smith visits their aunt for the holidays,” or “two hostile insurgent groups fired upon each other.” In the first example, the subject is a system and the object and individual, in the second example both subject and object are systems.
Where things get intricate
The ontological building blocks of subject-object disjunction, pairing, and system are sufficient for describing a wide range of social interactions, namely those that fall into category 1 and category 2b.
So what about category 2a, our interactions with non-human life forms? Well, its a grey area, and it also has many of the problems in common with our interactions with “artificial idiots.”
How we interact with non-human life can, in a simple view, be explained by how much non-human life resembles us. When we talk, at one end of the spectrum, about bacteria, we characterise them as entities that are completely independent of us, and that can, in essence, be treated as little machines. In other words, we usually use the subject-object disjunction when talking about them. The picture is very different at the other end of the spectrum. Think of apes, think of pets. We bestow (truncated) personalities to them, and especially pets are often seen as part of families. At this end of the spectrum, animals become part of our social fabric, and the subject-object disjunction is not longer adequate, animals become parts of our social groups. For this reason, human-animal interaction has actually been the subject of sociological research (see, for instance [Jerolmack 2005]). Other types of interactions are of course also occurring, for instance the interaction of groups of people and groups of animals. This kind of view and interaction is, for instance, at the centre of animal-conservation activities.
The above “grey scale” between life forms as mere objects and system theory (life forms as part of our social fabric) is a traditional heuristics that seems to work well. However, at second sight, it is rather makeshift and quite problematic. Let me elaborate on these shortcomings since the insight gained will come handy when discussion our interaction with “artificial idiots”.
Bacteria r us
When, for instance, addressing our relationships with micro-organisms outside the human body, e.g. bacteria in the stratosphere [Yang 2008], the use of a subject-object disjunction in combination with a system perspective of bacteria biotopes is rather appropriate, since these micro-organisms do not directly interact with us. In other words, the interactions are weak and indirect. However, the picture changes entirely when addressing the human microbiome, i.e. the micro-organisms living in and upon the human body [Wikipedia 2014b].There are actually ten times more micro-organisms in the microbiome than cells in the human body [Blaser 2013], and changes in the microbiome can have grave consequences for human health [Foxman 2013]. Our understanding of the microbiome is still quite limited, and it is not yet clear, if a disjunction of microbiome and human body actually makes sense or whether the unity of both have to be seen as a super-organism, as a new system [Foxman 2008].
So, even when we talk about extremely simple life forms, the subject-object boundary between humans and micro-organisms can be rather blurred.
Objectification and animal cruelty
As pointed out above, domestic animals are often seen as an extension of our social groups (dogs as members of the family), but what about farming animals ans wild animals? Both are usually objectified, and it has been argued that the prevalence of animal cruelty in factory farming is -partially- explained by our detached, emotionless view of animals as machines [Foer 2012]. However, this view of non-domestic animals was challenged already in the 20th century, and the challenge is ongoing [Duncan 2006]. One of the main arguments for being less cruel to animals rests on the observation that the arbiter for cruelty should not be how closely we socially relate to animals, rather whether they can experience pain and whether they suffer. So what has changed here is not so much a blurring of the subject-object disjunction, as in the case of the microbiome, but a revised picture of what animals are and what our relationship to other species should be contingent upon (suffering).
It is my contention that our interactions with “artificial idiots” belongs to both models of description. On the one hand we have the human-machine super-organisms which emerges from our interactions with “artificial idiots” (think human and microbiome). On the other hand, we will treat “artificial” idiots as mere subjects, and, as is the case for non-human animals, the potential for mistreatment is rather larger due to our oft limited understanding of “artificial idiots” and their interactions with their interactions. But there is more to this. Let me explain.
Why “artificial idiots” are different
It is my contention that our interactions with “artificial idiots” fall into a different category than the ones discussed so far. In order to make my case let me address traditional computers first.
Undoubtedly, the advent of computers has had a noticeable impact on humanity, and the same holds for the advent of the Internet, i.e. networks of connected computers. But one can make the point that even the Internet falls under category 1 (inanimate objects). Yes, the Internet is complex, and I do not think we understand it as a whole, but so far it has been gifted with only little autarky, i.e. the ability to generate its own rules. While a smart phone is many, many more times more complex than a Swiss knife, the smart phone can still be said to exhibit pre-defined behavioural patterns (maintain connection to the cellular network; poll text messages; issue low-battery alert …). It is this lack of decision autarky that justifies grouping (networked) computers under category 1. What changes though when (networked) computers are bestowed with the capability not only to make their own decisions but also to develop new decision rules? I would say: everything. Such computers are agents, and these agents will soon approach the complexity level of animals, and they will also operate as groups. So, instead of just seeing them as an object, one should probably see them as a social entity by themselves. However, when it comes to “artificial idiots,” even the meta-model of systems falls short of reality. What emerges are super-organisms of a different kind than in the case of the humans and the microbiome and humans and non-human animals.
Picture, for instance, semi-autonomous a distributed sensor network that helps us with monitoring a conservation area. If the network was not able to develop its own rules it would be appropriate to model it as a sensory extension of ourselves, akin a sophisticated camera. However, if the network, during its mission, can evolve and hone its decision rules, and maybe even react to unforeseen events (oil spill), the subject-object disjunction between us and the intelligent agents breaks down, since the autarkic actions of the sensor network have a direct influence on our thinking and behaviour. The agency of the network complements our own. The picture becomes even more blurred, when this sensor network also possesses actuating functionality, for instance the capability to regulate the water level in the conservation area. If the network did not possess agency it could be seen as a lever, a simple extension of ourselves, and agency changes this picture.
We already saw in the animal example that we often are obliged to change our understanding of “the other,” but applying a similar approach to “artificial idiots” will, in my opinion, not suffice, since in case of “artificial idiots” there is no there there. In contrast to farming animals and the like, “artificial idiots” will directly influence our thinking and they will also extend our actions, but in a somewhat autarkic way.
In principle, this situation is not new, we experience it daily in our interactions with other human beings. They posses agency, they learn, they develop high-level decision patterns, they share information with us, and their decisions influence us (and vice versa). So why not apply the same patterns to “artificial idiots?” The problem is the specificity of our social patterns, they are 100% honed for our interactions with humans. As soon as “the other” is not human, we resort to a subject-object disjunction (we exterminate termites; we let bees collect honey for us; we catch fish; we use computers; we drive cars …). But that resort is inadequate, since it totally neglects the epiphenomenon of the human-artificial system, which is constituted of entities with very different properties.
In summary, due to the strong connectedness between “artificial idiots” and humans, due to the fact that “artificial idiots” will influence our behaviour somewhat akin to how social interactions change our behaviour, and due to the fact that the actions of “artificial idiots” will not resemble human behaviour at all, what we will need is a new kind of sociology, a sociology that includes interactions between humans and artificial agents (and of course vice versa). This will not be a simple extension of sociology, since sociology assumes interaction between counterparts, while “artificial idiots” are anything but human. In other words, I hold that we do not have a scientific language for this symbiosis, and I am rather sceptical that a simple extension of existing modelling patterns will get us there. We seem thus to find ourselves in something akin an epistemic prison.
In the next blog post I will address how we might escape from our narrative prison (previous blog post) and the epistemic prison described here.
[Blaser 2013] Martin Blaser, Peer Bork, Claire Fraser, Rob Knight, and Jun Wang, “The microbiome explored: recent insights and future challenges,” Nature Reviews Microbiology, Vol. 11, No. 3, pp. 213-217, 2013.
[Bunge 2004] Mario Bunge, “How does it work? The search for explanatory mechanisms,” Philosophy of the social sciences, Vol. 34, No. 2, pp. 182-210, 2004.
[Duncan 2006] Ian J. H. Duncan, “The changing concept of animal sentience,” Applied Animal Behaviour Science, Vol. 100, No.1, pp. 11-19, 2006.
[Foer 2010] Jonathan Foer, Eating animals, Penguin UK, 2010.
[Foxman 2008] Betsy Foxman, Deborah Goldberg, Courtney Murdock, Chuanwu Xi, and Janet R. Gilsdorf, “Conceptualizing Human Microbiota: From Multicelled Organ to Ecological Community,” Interdisciplinary Perspectives on Infectious Diseases, Vol. 2008, Article ID 613979, 5 pages, 2008.
[Foxman 2013] Betsy Foxman and Mariana Rosenthal. “Implications of the Human Microbiome Project for epidemiology” American journal of epidemiology, Vol. 177, No. 3, pp. 197-201, 2013.
[Jerolmack 2005] Colin Jerolmack, “Our Animals, Our Selves? Chipping Away the Human–Animal Divide,” Sociological Forum, Vol. 20. No. 4, pp. Pp 651-660, 2005.
[Robinson 2012] Howard Robinson, “Dualism”, The Stanford Encyclopedia of Philosophy, Edward N. Zalta (ed.), available: http://plato.stanford.edu/archives/win2012/entries/dualism/, 2012.
[Wikipedia 2014a] Wikipedia contributors, “Phenomenology (philosophy),” Wikipedia, The Free Encyclopedia, http://en.wikipedia.org/w/index.php?title=Phenomenology_(philosophy)&oldid=616579660 (accessed August 15, 2014).
[Wikipedia 2014b] Wikipedia contributors, “Human microbiome,” Wikipedia, The Free Encyclopedia, http://en.wikipedia.org/w/index.php?title=Human_microbiome&oldid=617417386 (accessed August 15, 2014).
[Yang 2008] Yinjie Yang, Shiho Itahashi, Shin-ichi Yokobori, and Akihiko Yamagishi, “UV-resistant bacteria isolated from upper troposphere and lower stratosphere,” Biological Sciences in Space, Vol. 22, No. 1, pp. 18-25, 2008.
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.