FoCAS@ECAL Workshop: Steering Complex Adaptive Systems & Fundamentals of Collective Adaptive Systems

ECAL 2015 was the 13th European Conference on Artificial Life (ECAL 2015) held in York, UK, from 20-24 July 2015, hosted by the York Centre for Complex Systems Analysis at The University of York. The FoCAS Coordination Action joined forces with researchers from the complex systems community to run a workshop entitled Steering Complex Systems which presented a series of themed short invited talks with structured discussion.

Organised by Alexandra Penn (University of Surrey), James Dyke (University of Southampton), and Emma Hart (Edinburgh Napier University), the workshop was held on 20 July and considered the engineering challenges relevant to complex and collective adaptive systems. Conventional approaches to working with CAS are, for the most part, “brute force”, attempting to effect control in an input and effort intensive manner and are often insufficient when dealing with their inherent non-linearity and complexity. Such systems by their very nature are dynamic, adaptive and resilient and require management tools that interact with dynamic processes rather than inert artefacts. “Steering” in which we continuously interact with systems, is one way by which this might be accomplished, manipulating them or their environment via effective leverage points which exploit their structure and dynamics; monitoring their response and responding to their adaptation. However, the plethora of tools and techniques plus the overarching methodological framework required for this approach is at a nascent stage.

This was a broad-ranging and discursive workshop aiming to identify key ideas within, and the implications for, new paradigms.

Video: Highlights of the workshop discussion

Guest Speakers

SETTING THE SCENE: OVERARCHING QUESTIONS

Sarah Cornell (Stockholm Resilience Institute)
“Sustainability challenges in a complex world”.
Abstract
In the midst of calls from all sides to “navigate the Anthropocene”, it is probably good to pause and consider what would actually be involved in steering society for global sustainability. The most prominent calls often demand fundamental, not incremental, transformations – of the global polity, the economy, of
sectors, of mindsets (and usually of others’, rather than our own). They often reflect an immense wealth of information about societal and environmental change, and at the same time, they often demonstrate a stark poverty of insight about what “steering” complex and linked social and biophysical systems might
entail. Rather than focusing on the essence of sustainability – harmony among humans, and between humans and nature, to use the terms of the 1987 Brundtland Report – I will frame it as a simple knowledge/action dialectic, because this is one way to explore just how tangled, fragmentary and sometimes contradictory our notions are of systemic complexity, adaptive capacity, and purposive change when we are talking about people’s place in our living environment. These tangles and muddles are evident in both research and policy relating to global change. And of course most people think and do things despite research findings and policy interventions. The nub of my question, then, is what can be done in light of the apparently universal tendency to complexify when things are simple and simplifywhen things are complex.

Vivek Nallur (Trinity College Dublin)
“Where shall we have lunch?” Problems for a computer-aided future.
Abstract
Cities of the future will be filled with technology permeating every aspect of citizens’ lives. From decisions about personal health, to neighbourhood energy use, to city-wide decisions about transportation mechanisms, our strategies (indeed policies) will be influenced by technical systems that are autonomously adapting to individual and collective needs. Trusting such hugely complex adaptive systems will require being able to answer some fundamental questions, about systems that no single human is fully able to comprehend. Would we trust a system’s decision to re-arrange transport schedules of an entire city, for instance? Would we trust it merely because it is ‘intelligent’? Or would we require proof? What happens when multiple groups of humans, with multiple, conflicting goals request for resources? Is the system-optimum more important than social equity or fairness? Who is in charge? There are no definitive answers to these questions, but any answer will have to incorporate concepts of learning about changing goals, diversity, emergence and irrationality, all of which are not principal concerns of system design currently. Our current conception of software design focusses on the most efficient, elegant way to create systems, with as little waste as possible. Concepts like diversity stand in direct opposition to such design philosophies, by tolerating (or even encouraging) redundant components, inefficient architectures, sub-optimal algorithms, etc. When systems start interacting with each other, their interactions could give rise to unforeseen collective behaviour such as most obviously seen in flocking birds (and traffic jams). Such behaviour, called emergence, can have cascading consequences for systems with feedback loops. What does diversity mean for computer science, how can we predict/detect emergence, are questions that this talk will address.

EXAMPLES AND TOOLS: TECHNOLOGICAL AND BIOHYBRID SYSTEMS

Thomas Gabor (Ludwig-Maximilians-Universitat Munchen)
“The Liquid Computing Paradigm”.
Download Abstract (PDF)

Rob Mills (BioISI, Faculty of Sciences, U. Lisboa)
“On manipulating attractors in collective behaviours of bio-hybrid societies with robot interactions”.
Abstract
Bio­hybrid systems comprise animals and robots, the latter being accepted by the animals as part of that society. These are developed to study collective behaviours, making use of the robotic part to modify or interrogate certain behaviours in the animals. A landmark study by Halloy et al [1] showed cockroaches to socialise in part­robot groups in the same way as with all­cockroach groups. Moreover, when the robots’ environmental preferences were re­programmed to differ from the natural organism, they were able to induce aggregations in locations that were normally preferred less. These techniques offer much potential: the robots were able to override the natural behaviours of the cockroaches, despite being in the minority numerically; and note that this does not create a totally new behaviour, but rather, expresses an existing attractor in the space of social behaviours, in modified circumstances. Could this be generalised into dynamic reconfiguration of the bio­hybrid system, meaning that the robots would be able (within reasonable time) to repeatedly have the bio­hybrid society leave one attractor and get to another one? For instance, to guide a given swarm activity in time or space in order to avoid danger to the animals or other human activity? Our current research program within the ASSISI|bf project [2] also involves the development of bio­hybrid societies that contain multiple (natural) species that do not usually share habitats. Here, we must overcome challenges relating to sufficient communication of cues between the animals via the robots. Could this be applied in assisting ecosystems that have been disrupted by mankind? Is it possible to substitute the animals that fill given niches or roles that have been eroded, by introducing novel inter­species interactions? If an original species can no longer cope with a changed environment, would it be possible to “transfer” the role to a novel species that would not usually be able to interact appropriately, but through robot­mediated communication, have access and influence over sufficient aspects to maintain the niche and hence overall ecosystem function?
[1] Halloy et al, Science 318:1155 (2007)
[2] http://assisi­project.eu

NEW APPROACHES: SOCIOLOGICAL AND PHILOSOPHICAL PERSPECTIVES ON MANAGING LIVING SYSTEMS

Anna Krzywoszynska (University of Durham)
“Uncertainty, intuition and care in the management of vineyards and wine fermentations”.
Abstract
The Enlightenment taught us to make (or at least, to claim to be making) decisions about what we should do, now and in the future, on the basis of reliable knowledge about the past. This approach has led to the dominance of such tools as statistics, cost benefit analysis and risk assessments, which articulate our implicit belief that the past is knowable, and that this knowledge is an indication of the future. As this workshop demonstrates, we are slowly moving away from a conception of the world as fully knowable, and towards the challenge of acting responsibly in a world which does not hold still to allow for its accounting, but which is indeterminate, complex, and adaptive. In the light of this, I invite the participants of the workshop to explore three provocative questions – What is the link between knowledge and action? (How much) do we need to know in order to act? And what counts as actionable knowledge?

Simon McGregor (University of Sussex)
“Wrangling Complex Systems – A Near-Life Perspective On Complex Systems Control”.
Abstract
The human species owes our peculiar evolutionary success to at least two cognitive abilities. The first is an unparalleled ability to construct, understand and exploit simple mechanical systems. The second is a sophisticated ability to anticipate and interact with the behaviour of other living agents, including our social peers.

It appears that we use complementary strategies in these two domains. Mechanical systems are understood in terms of physical cause­effect relations, whereas agents are understood in terms of beliefs, desires and rationales. These strategies correspond to what the philosopher Daniel Dennett has called, respectively, the physical stance and the intentional stance.

Problems can arise when the systems we wish to understand (or interact with) fall between these two stools: too complex to characterise mechanically, but not “animate” enough to invite an intentional stance. I will propose that we should not abandon the reductive mechanical strategy in understanding such systems, but we should also begin to make wider use of biological analogies, and our innate capacity to comprehend systems in terms of their motives
and goals.

As such, I will argue that many complex systems (including both physical phenomena such as the weather, and social systems such as economies) fall into the category of near­life phenomena: displaying some of the characteristic properties of living systems, without being composed of cells, proteins or biological material.

We will see that there are continuities between far­from­equilibrium phenomena in cognitive neuroscience, cell biology, chemistry and even astrophysics; similarities which go beyond well­known thermodynamic generalities, by encompassing behaviours such as inference, self­replication and goal­directed locomotion. Additionally, I will illustrate how mathematical treatments of cognition can be applied to such systems, so that there is a rock­solid scientific
sense in which they can be attributed beliefs and desires.

With these continuities in mind, I will argue that a number of historical and cultural factors have held modern scientists back from properly engaging with the life­like nature of complex non­biological systems. Perhaps instead of assuming that we can steer such systems like the pilot of a mechanical vehicle, we should also start to think about what the systems “want” and “think”, so that ­ where appropriate ­ we can exploit our animistic intuitions successfully, in addition to our mechanistic ones.

ERIEUniversity of SurreyFoCAS


The Evolution and Resilience of Industrial Ecosystems (ERIE) is an EPSRC “Complexity Science for the Real World” interdisciplinary project. Linking collaborators from the departments of Sociology, Maths, Computing and The Centre for Environmental Strategy at the University of Surrey, ERIE applies complexity science to socio-economic systems to provide practical tools for decision makers and stakeholders managing large-scale human ecosystems.

FoCAS brings together researchers active in Collective Adaptive Systems. This FP7 coordination action is funded by Future and Emerging Technologies at the European Commission.


ARCHIVE


FoCAS-EriE Workshop @ECAL2015

At the European Conference on Artificial Life 2015, University of York

http://www.cs.york.ac.uk/nature/ecal2015

When: Monday 20th July 2015
Time: 10:00-13:00

*** Download timetable *** (PDF)

Organised by Alexandra Penn (University of Surrey) and James Dyke (University of Southampton), Emma Hart (Edinburgh Napier University) and Ben Paechter (Edinburgh Napier University)

The potential of new technologies which emulate or exploit unique properties of living or adaptive systems is widely lauded. Such technologies however, create new engineering challenges which must be addressed before they can become broadly utilised. Additionally, many pressing challenges for society today are inherently concerned with gaining a better ability to understand and manage interacting living or life-like systems upon which we rely. Problems in these areas demand a better ability to manage complex adaptive systems (CAS) than is currently available.

Socio-technical adaptive systems are often comprised of extremely large numbers of different units, each of which may have individual properties, objectives and actions. Boundaries between or within CAS can be fluid and the units can operate at different temporal and spatial scales often with conflicting goals. Decision-making is usually distributed and possibly highly dispersed, and the interaction between units may lead to the emergence of unexpected phenomena. Understanding the mechanisms that underpin the design and operation of CAS systems poses significant challenges.

Conventional approaches to working with CAS are, for the most part, “brute force”, attempting to effect control in an input and effort intensive manner and are often insufficient when dealing with their inherent non-linearity and complexity. Such systems by their very nature are dynamic, adaptive and resilient and require management tools that interact with dynamic processes rather than inert artefacts. “Steering” in which we continuously interact with systems, is one way by which this might be accomplished, manipulating them or their environment via effective leverage points which exploit their structure and dynamics; monitoring their response and responding to their adaptation. However, the plethora of tools and techniques plus the overarching methodological framework required for this approach is at a nascent stage.

Importantly, many of the complex systems which we would most like to influence have significant social components and may require the integration of participatory or political processes with tools from complexity science. Accomplishing this effectively will rely on interdisciplinary efforts encompassing social and political, as well as natural, sciences, engineering and philosophy.

This will be a broad-ranging and discursive workshop aiming to identify key ideas within, and the implications for, new paradigms. A range of guest speakers will set the scene with short talks pulling out key themes and questions from their different perspectives and domains of focus. Structured discussion will attempt to identify key issues, opportunities and challenges in steering complex adaptive systems.