Both humans and nonhumans can commit to participate in distributed problem solving in smart systems. Therefore the state of the art in collaborative coordination in agent-based smart systems, commitment to joint action, and the potential dysfunctional cooperative behaviour in such social computing systems is described.
Software agents are no longer mere tools, but have become interactions partners. The degrees of freedom built into computational artefacts can materialize in individual acts, mandated actions or collaborative interaction. New capabilities may emerge over time on the individual level. Self-organisation and coalition forming on the group level can occur. New cultural practices and novel institutional policies may emerge. Due to these developments we may speak of a participatory turn when assessing the current division of labour between humans and nonhumans.
Participation of human (and nonhuman) actors in computer-based environments requires the communicative involvement within a computer-mediated and (frequently) open organisational structure where a predefined goal is pursued.
Purely human online participation is explored in a wide variety of research projects
e.g. at the Alexander von Humboldt Institute for Internet and Society [
Participation of nonhumans (and humans) can be found in multiagent systems
(MAS). MAS focus on the simulation of complex interactions and relationships of
individual human and/or nonhuman agents. They represent a variant of social
computing systems. Examples range from swarm intelligence systems to the simulation of
sophisticated organisational structures. Social computing systems and especially MAS
may be deployed in experimental environments as well as outside the laboratory. In
testbed environments they are composed exclusively of software agents. From a
computer scientist’s perspective they are best suited to offer heuristics for NP-complete
problems in planning, optimization and all kinds of knowledge acquisition in open
environments where knowledge is local and distributed. They represent a variant of
crowd-based socio-cognitive systems (CBSC). As Pablo Noriega rightly remarked
after the workshop CBSC may also enable interactions to accomplish activities that
need not (may not) be conceived as problems and even when you design one such
system to solve one particular problem there needs not be an epistemic challenge.
While this is also true for MAS, it must be noted that they are currently mainly used
in computational sciences projects - may it be in computational science and
engineering, computational sociology or even in legal engineering: “Crowd simulation”
systems are useful if evacuation plans have to be developed. Demonstrators for the
coordination of emergency response services in disaster management systems, based on
electronic market mechanisms, have been built [
Moreover, MAS provide a basis to cyberphysical systems. Whereas “classical
computer systems separate physical and virtual worlds, cyberphysical systems (CPS)
observe their physical environment by sensors, process their information and influence
their environment with actuators according to communication devices” [
The individual elements of smart systems may be defined as “miniaturized devices
that incorporate functions of sensing, actuation and control. They are capable of
describing and analyzing a situation, and taking decisions based on the available data in
a predictive or adaptive manner, thereby performing smart actions. In most cases, the
“smartness” of the system can be attributed to autonomous operations based on closed
loop control, energy efficiency, and networking capabilities” [
These systems form part of the intelligent infrastructure of today’s world. Smart systems have a huge impact on our socio-cognitive environment since „machines don’t just replace what we do, they change the nature of what we do: by extending our capabilities, they set new expectations for what’s possible and create new performance standards and needs. …Our tools change us” [10, p.5]. Moreover it can be stated that in systems where humans and nonhumans collaborate “we’ll outsource some decisions to machines completely, while also assimilating computational rationality into our own decision processes” [10, p.2]. To put it more precisely: “the delegation of control functions to autonomous machines limits the options for human actions and decisions thus increasingly forcing humans into adaptive behaviour” [11, p.28]. Even such adaptive behaviour is a nontrivial task since these systems may be able to adapt to changes in the environment themselves. One option for potentially successful interaction and coordination of humans and nonhumans is offered by the above mentioned multiagent systems.
Current agent-based software systems range from swarm intelligence systems, based
on a bionic metaphor for distributed problem solving, to sophisticated e-negociation
systems [
Higher degrees of freedom are provided if the commitment to a specific task or even
to distributed problem solving is not fixed during execution but may change. In the
human case “commitments and predictability in joint action” are a research field in its
own right (u. a. [
However, the fact that current technical agents “lack humans’ consciousness,
intentionality and free will” (Moor 2006, p. 20) does not mean that they do not possess a
degree of “social autonomy in a collaborative relationship”. This form of
goalautonomy was defined by Falcone and Castelfranchi as having to two components:
“a) meta level autonomy that denotes how much the agent is able and in condition of
negotiating about the delegation or of changing it; b) a realization autonomy that
means that the agent has some discretion in finding a solution to an assigned problem,
or a plan for an assigned goal” [17, p. 407]. Even certain current software agents may
possess this kind of social autonomy thus displaying a certain proto-social behaviour.
Such software agents need not necessarily be based in a Belief-Desire-Intention
(BDI)-model [17, p. 416]. However, if one intends to base a computational model of
trust on BDI-agents, an elaborate approach is to be found in [
This paper cannot expand on the similarities and differences of current human and technical agents. It must suffice to state that human capabilities and those of technical agents may differ widely. Their acts are based on different cognitive systems, different degrees of freedom and only partially overlapping spheres of experience.
Even criminal behaviour, deliberate misinterpretations of norms or negligence can be
studied in MAS if it is based on bounded rationality. Investigations into machine
ethics and the treatment of artificial agents as legal subjects are very instructive when
searching for commonalities and fundamental differences in unethical or illegal
behaviour between humans and nonhumans. Books as “the law of robots” [
Chopra and White are convinced that “in principle artificial agents should be able to qualify for independent legal personality, since this is the closest legal analogue to the philosophical conception of a person” [21, p. 182]. In their view “artificial agents are more likely to be law-abiding than humans because of their superior capacity to recognize and remember legal rules” [21, p. 166]. If they do not abide by the laws “a realistic threat of punishment can be palpably weighed in the most mechanical of cost-benefit calculations” [21, p. 168].
Pagallo perceives the legal personhood of robots and their constitutional rights as an
option only being relevant in the long term [20, pp. 147]. However he discusses at
length both human greediness using robots as criminal accomplices and artificial
greediness. He states that “in certain fields of social interaction, “intelligence”
emerges from the rule of the game rather than individual choices” [20, p.96]. Thus such
social and asocial intelligence might be acquired by (rational) nonhuman agents, too.
Moreover investigations into the potential ethical status of software agents have been
undertaken (e.g. [
In order to clarify the state of the art in software agents’ ethics Moor’s distinctions
between ethical-impact agents, implicit ethical agents, explicit ethical agents and full
ethical agents may be used [
Analogous to this classification of ethical behaviour displayed by software agents a wide variety of amoral agents could be implemented. They could range from of unethical impact agents, implicit unethical agents to explicit unethical agents e.g. based on virtue ethics. They could be modelled for use in online games. Such games could provide sheer entertainment, edutainment or form part of the currently so popular serious games. The latter “have an explicit and carefully thought-out educational purpose and are not intended to be played primarily for amusement” [24, p.5]. Agent-based models allow to model a wide variety of social and asocial behaviour. Yet when transferring the insights gained in the laboratory to real world scenarios, one must proceed with great care. Humans, even if they do not always “follow the rules of the game” are able to perceive others not only as social tools but as valuable peers and act accordingly.