Search

Complex adaptive systems composed of self-interested agents can in some circumstances self-organise into structures that enhance global adaptation or efficiency. However, the general conditions for such an outcome are poorly understood. In contrast, sufficient conditions for artificial neural networks to form structures that perform collective computational processes such as associative memory/recall, generalisation and optimisation, are well-understood. While such global functions within a single agent or organism may arise from mechanisms (e.g., Hebbian learning) that were selected for this purpose, agents in a multi-agent system have no obvious reason to produce such global behaviours when acting from individual interest. However, Hebbian learning is actually a very simple and fully-distributed habituation or positive feedback principle. Here we use an adaptive network model in which agents can modify their behaviours (states) but also their interactions with other agents (network topology). We show that when self-interested agents can modify how they are affected by other agents then, in adapting these inter- agent relationships to maximise their own utility, they will necessarily alter them in a manner homologous with Hebbian learning. When the agents adapt their behaviours relatively quickly, and their relationships with other agents relatively slowly, we find that the overall network dynamics are modified to find better adapted states more reliably. This separation in timescales causes the state dynamics to spend most of their time at attractors. Thus, the network develops an associative memory that amplifies a subset of its own attractor states. This self-organised modification to the network dynamics enhances its ability to resolve conflicts between agents. Moreover, we show that the system is not merely ‘recalling’ high quality states that have been previously visited, but ‘predicting’ their location by generalising over local attractor states that have already been visited. Thus, globally adaptive behaviours can emerge from self-organising adaptive networks that follow organisational principles familiar in connectionist models of organismic learning.

Document is itself an extended abstract.

Document is itself an extended abstract.

The pattern of gene expression in the phenotype of an organism is determined in part by the dynamical attractors of the organism’s gene regulation network. Changes to the connections in this network over evolutionary time alter the adult gene expression pattern and hence the fitness of the organism. However, the evolution of structure in gene expression networks (potentially reflecting past selective environments) and its affordances and limitations with respect to enhancing evolvability is poorly understood in general. In this paper we model the evolution of a gene regulation network in a controlled scenario. We show that selected changes to connections in the regulation network make the currently selected gene expression pattern more robust to environmental variation. Moreover, such changes to connections are necessarily ‘Hebbian’ – ‘genes that fire together wire together’ – i.e. genes whose expression is selected for in the same selective environments become co-regulated. Accordingly, in a manner formally equivalent to well-understood learning behaviour in artificial neural networks, a gene expression network will therefore develop a generalised associative memory of past selected phenotypes. This theoretical framework helps us to better understand the relationship between homeostasis and evolvability (i.e. selection to reduce variability facilitates structured variability), and shows that, in principle, a gene regulation network has the potential to develop ‘recall’ capabilities normally reserved for cognitive systems.

The Community of Practice/Community of Learning approach facilitates networks of people who have common interests and aims. It is through these communities, both informal and formal, that participants are able to interact, communicate, learn from one another, solve problems and create new knowledge. While the concept is certainly not new, many of the recent developments have seen organisations seeking to harness and develop knowledge that ?adds value? through knowledge management, individual and organisational learning. The central aim of this feasibility study was to identify the main issues that would influence the design and development of a Community of Pedagogic Practice (CoPP) at Southampton Solent University. There were four main parts to the study, starting with a detailed review of the literature relating to Communities of Practice/Learning Networks and Virtual communities of practice. Four leading practice Higher Education Institutions, with recently launched and/or established communities of learning were identified. Interviews with key members of staff from these institutions were used to gain an in-depth understanding of their experiences in developing and maintaining communities of learning. Internal stakeholder research at Southampton Solent University was also completed this included one to one interviews with key stakeholders responsible for research and scholarship, staff development, and learning and teaching. An online survey was administered in order to establish current and likely future pedagogic learning preferences of all academic and support staff. Finally, the technological assessment sought to evaluate the potential of the University intranet to provide an online space to support the development of a community of practice. The study concluded that there is broad support at Southampton Solent University for the establishment of a Community of Learning to support pedagogic research and share effective pedagogic practice. Similar initiatives at other Higher Education Institutions are proving beneficial and attracting considerable interest (for example the Inquiring Pedagogies (iPED) Research Network at Coventry University). Experience shows that the development of a successful Community of Learning will require time and persistence. A number of barriers and areas for particular attention have been identified from the literature and case study institutions. Awareness of these issues by the leaders of the proposed Community of Learning as Southampton Solent University will reduce the time needed for the network to become established and increase the return on investment.

In neural networks, two specific dynamical behaviours are well known: 1) Networks naturally find patterns of activation that locally minimise constraints among interactions. This can be understood as the local minimisation of an energy or potential function, or the optimisation of an objective function. 2) In distinct scenarios, Hebbian learning can create new interactions that form associative memories of activation patterns. In this paper we show that these two behaviours have a surprising interaction – that learning of this type significantly improves the ability of a neural network to find configurations that satisfy constraints/perform effective optimisation. Specifically, the network develops a memory of the attractors that it has visited, but importantly, is able to generalise over previously visited attractors to increase the basin of attraction of superior attractors before they are visited. The network is ultimately transformed into a different network that has only one basin of attraction, but this attractor corresponds to a configuration that is very low energy in the original network. The new network thus finds optimised configurations that were unattainable (had exponentially small basins of attraction) in the original network dynamics.

We present a model that considers evolvable symbiotic associations between species, such that one species can have an influence over the likelihood of other species being present in its environment. We show that this process of 'symbiotic evolution' leads to rare and adaptively significant complexes that are unavailable via non-associative evolution.

Symbiosis, the collaboration of multiple organisms from different species, is widespread amongst prokaryotes. As symbiotic associations become more absolute, an inseparable union can result (symbiogenesis). This union would typically be considered a new unit of selection once a set of symbionts became reproductively inseparable. However we consider a macroevolutionary model within which new units of selection are formed gradually from maturing symbiotic relationships, without the need for symbiogenic events to be complete. We find that these adaptive units of selection can evolve complexes that are provably unevolvable under fixed units of selection. We perform experiments in several structured rugged landscapes in which there only few global optima amongst many local optima. Importantly, the number of initial conditions that must be sampled to find one of these global optima by individual adaptation is exponentially large (i.e., their basins of attraction are small). Selection on individual species only has the ability to evolve to locally optimal configurations. Selecting on groups at the level of entire ecosystems requires an exponentially large number of ecosystems to be created to find optima of globally maximal fitness. Since this level of selection cannot utilise gradient information, global optima are only guaranteed to be found by enumerating all possible ecosystem configurations. In our model the ecosystem is sub-divided into several semi-independent demes, each of which evolves to local attractors via individual adaptation. Symbiotic associations between the entities that are present in the local attractor discovered by each deme are modified to reflect their frequency of co-occurrence in the attractors across all of the demes. These two phases are repeated, resulting in coalitions of increasing size, stability, and co-dependence. We compare two variants for how the symbiotic associations are formed: associations between entities that are ‘all-or-nothing’, and probabilistic associations. Selection starts at the level of the individual, when local optima are discovered. However, the unit of selection adapts as composites gradually develop. Some composites are favoured at the expense of other composites, leading to the discovery of higher fitness optima. These composites consist of co-adapted groups that are locally optimal under individual selection, and not random configurations of large sets of entities. Thus, although the groups can contain many entities, the number of competing complexes is small: the configuration space is reduced as symbiotic relationships are strengthened. We demonstrate the adaptive significance of these processes by applying them to provably difficult optimisation problems. In general, optimisation problems frequently have some structure that can be exploited, but only with an appropriate mechanism that can recognise that structure. When locally optimal configurations contain some information that is congruent with globally optimal solutions, such as when local optima are created from the frustration of a large number of low-order fitness interactions, the process described above can provide automatic problem decomposition. When composite entities evolve through the intensification of symbiotic coalitions, the units of selection adapt causing the configuration space dimensionality to be collapsed. We show that this can result in complexes that are unevolvable when the units of selection are static.

The expression level of a gene in future generations can be modified both by genetic mutations and by the attachment of methyl groups to the DNA. Since the DNA methylation pattern along a genome is inherited, methylation patterns constitute a significant epigenetic inheritance mechanism that is subject to evolution by natural selection. The variation rate of methylation patterns is generally higher than that of DNA which suggests that evolution of methylation patterns might be more rapid than that of genetic evolution. But, common consequences of methylation, such as reduced expression of methylated genes, could also be produced by genetic changes and these would have higher heritability. The question we address in this work is how the evolution of epigenetic methylation-dependent phenotypes might interact with the evolution of genetic DNA-determined phenotypes. There is no biological mechanism known to directly transfer methyl groups into equivalent DNA changes. However, in principle an indirect mechanism could cause evolved methylation patterns to enable the subsequent evolution of equivalent genetic patterns in a manner analogous to the Baldwin effect (Baldwin, Am. Nat., 30:441-451, 1896; Jablonka et al, TREE, 13:206-210, 1998). The Baldwin effect describes how non-heritable acquired characteristics can influence the evolution of equivalent genetic characteristics without any direct Lamarckian inheritance of acquired characters. This occurs because the ability to acquire or learn a new behaviour changes the selective pressures acting on genetic changes. Specifically, genetic changes that support this behaviour, e.g. by reducing learning time by making a small part of the behaviour genetically innate, may be selected for when the learning mechanism is present even though these same genetic changes may not be selected for when the learning mechanism is absent. Over generations, the modified selection pressures so produced can cause genetic assimilation of a phenotype that was previously acquired, even to the extent of making the acquisition mechanism subsequently redundant. Thus a learned behaviour can guide the evolution of an equivalent innate behaviour (Hinton & Nowlan, Complex Systems, 1: 495-502, 1987). In the Baldwin effect a rapid mechanism of lifetime adaptation guides the relatively slow genetic evolution of the same behaviour. By analogy, Jablonka et al have suggested that “genetic adaptations may be guided by heritable induced or learnt phenotypic adaptations”. Here we hypothesise that “inherited epigenetic variations may be able to ‘hold’ an adapted state for long enough to allow similar genetic variations to catch up”, as they put it, even if the epigenetic variations are not induced or learnt but simply evolved by natural selection on methylation patterns. We assume that an individual may only express one phenotype in its lifetime, but that a given genome will persist relatively unchanged on a timescale that allows its methylome to adapt by natural selection. Thus, in contrast to the Baldwin effect, in this case two mechanisms of evolution by natural selection are coupled — one acting at a different variation rate from the other. We present a simple model to illustrate how a rapidly evolving methylome can guide a slowly evolving but highly-heritable genome. This is used to show that methylome evolution can enable genetic evolution to cross fitness valleys that would otherwise require multiple genetic changes that were each selected against. This finding suggests that the relatively rapid evolution of methylation patterns can produce novel phenotypes that are subsequently genetically assimilated in DNA evolution without direct transfer or appeal to induced phenotypes. This can enable the genetic evolution of new phenotypes that would not be found by genetic evolution alone, even if methylation is not significant in the ultimate phenotype.

The alternative phenotype hypothesis contends that multiple phenotypes exist in a single genotype and are expressed by environmental or genetic cues. It further states that these multiple phenotypes will be maintained and improved in a population where the environment is unstable, in spite of the increased cost of this plasticity. In this work we propose a simple computational model to investigate the conditions under which alternative phenotypes become beneficial, and persist over evolutionary timescales. We find that the environment must vary to realise this hypothesis, and that these adaptations not only provide a fitness benefit in highly unstable environments but also continue to arise despite increasing stability and a corresponding gradual decline in fitness.

This paper is a policy review of the role of investment climate in post-conflict situations. It summarizes the broad range of ways in which conflict negatively affects the investment climate, from macroeconomic instability to a degraded regulatory framework. It stresses that attention needs to be paid to the broader"enabling environment,"including institutions, governance, capacity, and social capital. It suggests that a vibrant private sector underpinned by a good investment climate is particularly important in the post-conflict recovery phase for three reasons: it generates employment, provides public services where the state has retrenched, and builds social capital. By addressing these important"greed and grievance"factors, the private sector helps reduce the likelihood of a return to conflict. The paper concludes by distilling key lessons relating to the management of the post-conflict reform process. Despite the importance of a good investment climate, greater effort is needed to ensure that private sector development reforms are included in the first round of post-conflict policymaking. Local ownership of reforms and enhanced local capacity to implement them is key to sustainable improvements in the investment climate. Development partners have an important role to play in facilitating dialogue and promoting partnerships between public and private sector stakeholders. At the same time, development partners need to ensure that their presence in fragile post-conflict economies does not damage the very sector they are trying to support.

Escaping local optima and crossing fitness valleys to reach higher-fitness regions of a fitness landscape is a ubiquitous concept in much writing on evolutionary difficulty. The Baldwin effect, an interaction between non-heritable lifetime plasticity (e.g. learning) and evolution, has been shown to be able to guide evolutionary change and ‘smooth out’ abrupt fitness changes in fitness landscapes –thus enabling genetic evolution that would otherwise not occur. However, prior work has not provided a detailed study or analysis on the saddle-crossing ability of the Baldwin effect in a simple multi-peaked landscape. Here we provide analytic and simulation studies to investigate the effectiveness and limitations of the Baldwin effect in enabling genotypic evolution to cross fitness valleys. We also discuss how canalisation, an aspect of many prior models of the Baldwin effect, is unnecessary for the Baldwin effect and a hindrance to its valley-crossing ability.