The overtake car maneuver involves high risk and complex judgement. For autonomous vehicles this is challenging, especially for human-initiated overtake requests. If a user requests the maneuver there must be a rapid safety assessment. Language models have great potential to classify safety with explanations, but they struggle to disentangle critical information from complex vehicular environments. We apply CLEAR (Cue Learning using Evolution for Accurate Recognition) to evolve prompt cues that optimize the ability of language models to correctly predict safety scores for overtaking maneuvers. To achieve this, we create a novel open-source symbolic traffic model EvoDrive designed specifically for EC research, which outputs LLM-readable snapshots. We show that LLMs + EvoDrive with CLEAR can reduce error by more than 20% compared to without CLEAR, with statistically significant results. Analysis shows evolved cues are coherent and have reduced variability in LLM output.

Large Language Model (LLM) image recognition is a powerful tool for extracting data from images, but accuracy depends on providing sufficient cues in the prompt - requiring a domain expert for specialized tasks. We introduce Cue Learning using Evolution for Accurate Recognition (CLEAR), which uses a combination of LLMs and evolutionary computation to generate and optimize cues such that recognition of specialized features in images is improved. It achieves this by auto-generating a novel domain-specific representation and then using it to optimize suitable textual cues with a genetic algorithm. We apply CLEAR to the real-world task of identifying sustainability data from interior and exterior images of buildings. We investigate the effects of using a variable-length representation compared to fixed-length and show how LLM consistency can be improved by refactoring from categorical to real-valued estimates. We show that CLEAR enables higher accuracy compared to expert human recognition and human-authored prompts in every task with error rates improved by up to two orders of magnitude and an ablation study evincing solution concision.

Conceptual architecture involves a highly creative exploration of novel ideas, often taken from other disciplines as architects consider radical new forms, materials, textures and colors for buildings. While today's generative AI systems can produce remarkable results, they lack the creativity demonstrated for decades by evolutionary algorithms. SCAPE, our proposed tool, combines evolutionary search with generative AI, enabling users to explore creative and good quality designs inspired by their initial input through a simple point and click interface. SCAPE injects randomness into generative AI, and enables memory, making use of the built-in language skills of GPT -4 to vary prompts via text-based mutation and crossover. We demonstrate that compared to DALL. E 3, SCAPE enables a 67% improvement in image novelty, plus improvements in quality and effectiveness of use; we show that in just three iterations SCAPE has a 24% image novelty increase enabling effective exploration, plus optimization of images by users. We use more than 20 independent architects to assess SCAPE, who provide markedly positive feedback.

The use of autonomous robots for delivery of goods to customers is an exciting new way to provide a reliable and sustainable service. However, in the real world, autonomous robots still require human supervision for safety reasons. We tackle the real-world problem of optimizing autonomous robot timings to maximize deliveries, while ensuring that there are never too many robots running simultaneously so that they can be monitored safely. We assess the use of a recent hybrid machine-learning-optimization approach COIL (constrained optimization in learned latent space) and compare it with a baseline genetic algorithm for the purposes of exploring variations of this problem. We also investigate new methods for improving the speed and efficiency of COIL. We show that only COIL can find valid solutions where appropriate numbers of robots run simultaneously for all problem variations tested. We also show that when COIL has learned its latent representation, it can optimize 10% faster than the GA, making it a good choice for daily re-optimization of robots where delivery requests for each day are allocated to robots while maintaining safe numbers of robots running at once.

We inhabit an obscure planet, in an obscure galaxy, around an obscure sun, but on the other hand, modern human society represents one of the most complex things we know.—David Christian1If human societies are so complex, then how can we hope to understand them? Artificial Life gives us one answer. The field of Artificial Life comprises a diverse set of introspective studies that largely ask the same questions, albeit from many different perspectives: Why are we here? Who are we? Why do we behave as we do? Starting with the origins of life provides us with fascinating answers to some of these questions. However, some researchers choose to bring their studies closer to the present day. We are, after all, human. It has been a few billion years since our ancestors were self-replicating molecules. Thus more direct studies of ourselves and our human societies can reveal truths that may lead to practical knowledge.The articles in this special issue bring together scientists who choose to perform this kind of research. Expanded from submissions to our annual Agent-Based Modelling of Human Behaviour Workshop, the studies share similar methods, all using variations of agent-based modeling (ABM) to ask their own what-if questions. As guest editors, we believe such collections help bring together and enhance such research by sharing ideas. While ABM research—out of necessity—is often highly specialized toward the hypotheses and phenomena under study, the research methodology is shared by all. We formulate our hypothesis, develop our agent-based model of the relevant aspects of reality, and run experiments to gather evidence that may support or refute the hypothesis. An experimental model that supports the hypothesis may not prove that reality follows this approach or agrees with this result, but it indicates that there exists a specific set of conditions that, if found to be true elsewhere, may produce the same result. Modeling tells us about trends, about possible likelihoods. Our ABMs show us what will result if our assumptions are valid and why, whether we are examining civil violence, app stores, the economy, fish markets, language evolution, or energy consumption. When we study human societies, ABMs are the tools of choice for obvious reasons: It is not ethical or safe to play what-if experiments with ourselves. The researchers in this special issue demonstrate the exciting potential in ABM. We can create our own safe virtual worlds and make discoveries that enlighten us about ourselves.In this special issue, we have a fascinating collection of studies that relate to us. For example, from our first article, “The Effects of Information on the Formation of Migration Routes and the Dynamics of Migration” by Hinsch and Bijak, we may discover that as we spread across the planet both physically and culturally, the nature of our interactions affects us in ways we may not expect. The authors develop a spatially explicit agent-based model of human migration and show that information exchange plays a crucial role in affecting all aspects of migration: Too little information results in suboptimal migration routes, increased stochasticity, and migrants frequently not arriving at their preferred destinations, and too much information under certain conditions can lead to less predictable migration routes.Likewise, from our second article, “Adapting the Exploration–Exploitation Balance in Heterogeneous Swarms: Tracking Evasive Targets” by Kwa et al., we might recognize that like the robotic agents, we are diverse in every sense of the word, and that diversity is not only to be celebrated; it is a necessity for the optimal functioning of our societies. In their work, agents with different properties and behaviors are part of the same collective, forming a heterogeneous swarm. The authors use a decentralized search-and-tracking strategy with adjustable levels of exploration and exploitation and find that small, heterogenous swarms can match and outperform homogeneous swarms. They also explore using differentiated strategies to take advantage of heterogenous swarms.Our third contribution, “Expertise, Social Influence, and Knowledge Aggregation in Distributed Information Processing” by Mertzani et al., helps us understand that fields like psychology provide insights into human behavior and how we communicate—a topic close to our hearts as guest editors. Mertzani et al. develop an algorithm for knowledge aggregation based on Nowak’s regulatory theory of social influence. This theory posits that social influence consists not only of sources trying to influence targets but also of targets seeking sources by whom to be influenced and learning what processing rules those sources are using.Our societies face constant challenges that threaten us. Sadly, humans evolved to be violent and to work together for our violent aims, as ongoing geopolitical instability shows all too well. “Explaining Neuroevolution of Fighting Creatures Through Virtual fMRI” by Godin-Dubois et al. investigates where these behaviors come from and what causes different strategies and morphologies. The authors develop artificial creatures that engage in individual and team competitions and investigate three types of reactions—pain, vision, and audition—each with varying response levels but consistent across all evolution types.When we’re not battling each other, human societies also must battle pathogens—most recently in the global COVID-19 pandemic. As the virus becomes endemic and we develop coping strategies, are we responding appropriately to the changing risk levels? “Self-Isolation and Testing Behavior During the COVID-19 Pandemic: An Agent-Based Model” by Gostoli and Silverman models this phenomenon to find out. In this article, the authors develop an agent-based model that includes a behavioral module determining the agents’ testing and isolation propensity. According to the authors, most models focus on the replication of the interactions’ processes through which the virus is passed from infected agents to susceptible ones; in this model, agents modify their behavior as they adapt to the risks posed by the pandemic.Finally, there can be no more fundamental challenge to a society than its constant search for resources to sustain itself. Our final paper, “Social Search and Resource Clustering as Emergent Stable States” by Luthra and Todd, looks at the interplay between resources and consumers of those resources in a 2-D simulated world. Agents evolve strategies to search for food, and the resources resemble plants that grow continuously across the 2-D simulated world. Resources and other consumers produce distinguishable “odors” that are normally distributed around their positions in the grid, and agents can perceive these two types of odors and make movement decisions that take them toward or away from each kind of source.Through models such as these, we can learn why we might behave as we do as a species and whether we can start to mitigate or redirect some of the less-than-optimal patterns of behaviors to which our complex societies might be attracted. We hope you enjoy reading these articles as much as we have and that you might be inspired to ask your own what-if questions using ABM in your work.

Abstract Nature has spent billions of years perfecting our genetic representations, making them evolvable and expressive. Generative machine learning offers a shortcut: learn an evolvable latent space with implicit biases towards better solutions. We present SOLVE: Search space Optimization with Latent Variable Evolution, which creates a dataset of solutions that satisfy extra problem criteria or heuristics, generates a new latent search space, and uses a genetic algorithm to search within this new space to find solutions that meet the overall objective. We investigate SOLVE on five sets of criteria designed to detrimentally affect the search space and explain how this approach can be easily extended as the problems become more complex. We show that, compared to an identical GA using a standard representation, SOLVE with its learned latent representation can meet extra criteria and find solutions with distance to optimal up to two orders of magnitude closer. We demonstrate that SOLVE achieves its results by creating better search spaces that focus on desirable regions, reduce discontinuities, and enable improved search by the genetic algorithm.

Bridging the gap in software development from idea to need remains a difficult task for startups, especially when users are unreachable within the black boxes of large corporations. The lack of customer collaboration results in the failure of requirements elicitation and subsequently the failure of software products. In this work, we describe a simple but effective method to enable startups to reach relevant hidden stakeholders within large corporations and elicit requirements from them. AdvisorNet leverages professional networks and social media to find senior domain experts. These carefully selected advisors bridge the gap from the outside world to the hidden corporate structure and social networks in the target corporation and enable highly successful elicitation from stakeholders. We demonstrate AdvisorNet with a real-world case study in which a London-based startup successfully connected with multiple advisors who then introduced previously unreachable users for requirements elicitation and decision-makers to achieve their goal of trialing their product within the large corporations. Finally, we provide suggestions for future research to formalize the method further and enable it to become rigorous and repeatable.

Collaboration is fundamental to our society, but how should we best build teams? We investigate by applying optimisation to an agent-based model of collaboration. The model takes inspiration from particle swarm optimisation, abstracting a shared goal as a shared optimisation task, and modelling the personality differences in team members as strategies for moving within, interpreting and sharing information about the solution space. We update the model and add a genetic algorithm in order to investigate the effects of differing initial ideas within teams of different personality combinations as they attempt to collaborate to achieve their shared task. We run experiments on homogeneous teams with similar personalities and heterogeneous teams with random personalities and find that increased diversity in team members' initial ideas significantly improves teamwork, and more so for teams comprising individuals with similar personalities.

The emergence of Systems of Systems (SoSs) and Systems of Systems Engineering (SoSE) is largely driven by global societal needs including energy-water-food nexus, population demographics, global climate, integrated transport, security and social activity. However, due to their scale, structural and functional complexity and emergent properties, these global spanning Cyber-Physical Systems of Systems are becoming increasingly complex and more difficult for current requirements engineering (RE) practices to handle. In this paper, we firstly introduce SoSE as an emerging discipline and key characteristics of SoSs. We then highlight the challenges that the RE discipline must respond to. We discuss some weaknesses of current RE techniques and approaches to cope with the complexity of SoSs. We then argue that there is a need for the global RE community to evolve current RE approaches and to develop new ways of thinking, new RE capabilities and possibly a new RE science as a key mechanism for addressing requirements engineering complexities posed by Systems of Systems. We then outline a requirements engineering perspective and research agenda that identifies 'top-10' research themes informed by a cluster of four Systems of Systems Engineering projects funded by the European Commission's Horizon 2020 research programme.

Natural ecosystems exhibit complex dynamics of interacting species. Man-made ecosystems exhibit similar dynamics and, in the case of mobile app stores, can be said to perform optimization as developers seek to maximize app downloads. This work aims to understand stability and instability within app store dynamics and how it affects fitness. The investigation is carried out with AppEco, a model of the iOS App Store, which was extended for this paper and updated to model the store from 2008 to 2014. AppEco models apps containing features, developers who build the apps, users who download apps according to their preferences, and an app store that presents apps to the users. It also models developers who use commonly observed strategies to build their apps: innovator, milker, optimizer, copycat, and flexible (the ability to choose any strategy). Results show that despite the success of the copycat strategy, there is a clear stable state for low proportion of copycats in developer populations, mirroring results in theoretical biology for producer-scrounger systems. The results also show that the best fitness is achieved when the evolutionary optimizer (as producer) and copycat (as scrounger) strategies coexist together in stable proportions.

Mobile applications (apps) are software developed for use on mobile devices and made available through app stores. App stores are highly competitive markets where developers need to cater to a large number of users spanning multiple countries. This work hypothesizes that there exist country differences in mobile app user behavior and conducts one of the largest surveys to date of app users across the world, in order to identify the precise nature of those differences. The survey investigated user adoption of the app store concept, app needs, and rationale for selecting or abandoning an app. We collected data from more than 15 countries, including USA, China, Japan, Germany, France, Brazil, United Kingdom, Italy, Russia, India, Canada, Spain, Australia, Mexico, and South Korea. Analysis of data provided by 4,824 participants showed significant differences in app user behaviors across countries, for example users from USA are more likely to download medical apps, users from the United Kingdom and Canada are more likely to be influenced by price, users from Japan and Australia are less likely to rate apps. Analysis of the results revealed new challenges to market-driven software engineering related to packaging requirements, feature space, quality expectations, app store dependency, price sensitivity, and ecosystem effect.

App stores are one of the most popular ways of providing content to mobile device users today. But with thousands of competing apps and thousands new each day, the problem of presenting the developers' apps to users becomes nontrivial. There may be an app for everything, but if the user cannot find the app they desire, then the app store has failed. This paper investigates app store content organisation using AppEco, an Artificial Life model of mobile app ecosystems. In AppEco, developer agents build and upload apps to the app store; user agents browse the store and download the apps. This paper uses AppEco to investigate how best to organise the Top Apps Chart and New Apps Chart in Apple's iOS App Store. We study the effects of different app ranking algorithms for the Top Apps Chart and the frequency of updates of the New Apps Chart on the download-to-browse ratio. Results show that the effectiveness of the shop front is highly dependent on the speed at which content is updated. A slowly updated New Apps Chart will impact the effectiveness of the Top Apps Chart. A Top Apps Chart that measures success by including too much historical data will also detrimentally affect app downloads.

Due to an increasingly connected society and industry, our modern societal world and all industry sectors now increasingly depend on large-scale complex Systems of Systems (SoS). The emerging interdisciplinary area of SoS and Systems of Systems Engineering (SoSE) is largely driven by societal needs including public services such as health, transport, water, energy, food security, etc. The scale, complexities and challenges presented by SoS require us to go beyond traditional Requirements Engineering (RE) approaches. However, as is evident from publications in major Requirements Engineering conferences and journals, no significant effort has been expedited towards addressing specific RE issues for Systems of Systems Engineering. This panel explores key RE challenges in Systems of Systems Engineering, specifically, the areas in which the international RE community need to focus its research, and the approaches that are most likely to meet these challenges effectively. We first introduce Systems of Systems Engineering and outline key characteristics of SoS. We conclude by arguing that there is an urgent need for the global RE community to develop new ways of thinking, new capabilities and possibly a new science as a key mechanism for addressing requirements complexities posed by Systems of Systems.

The large-scale Systems of Systems (SoS) are there to improve the provision of societal services. These are driven by global drivers such as population growth, aging, food security, energy security and climate change. However, a considerable number of SoS projects resulted in failures, causing detrimental economic impact. This paper presents a set of preliminary results from an European Commission (EC) funded project called T-AREA-SoS (Trans-Atlantic Research and Education Agenda on Systems of Systems). The primary purpose of T-AREA-SoS is to provide input to EC's 80 billion Horizon 2020 research programme in formulating a Strategic Research Agenda for Systems of Systems Engineering (SoSE) that spans American and European activities. An initial set of prioritised SoSE research elements derived from an online survey are discussed from an economic and societal stance using examples from heterogeneous sectors such as transport, energy and manufacturing.

In mobile app ecosystems, an app can behave like a virus. Once downloaded, it may cause its user to recommend that app to friends who then may download the app and "infect" other friends. Epidemics occur when a small number of downloads causes a snowballing effect that results in a massive number of downloads (and consequently, a rich developer). This paper presents AppEco, the first Artificial Life model of mobile application ecosystems. AppEco models the app store, app developers, apps, users, and their behaviour. We use AppEco to simulate Apple's iOS app ecosystem and investigate common publicity strategies adopted by developers and their effects on app downloads. Specifically, we investigate three causal factors for a widespread "app infection" from epidemiology: the users' exposure to the app, the users' susceptibility to the app, and the infectiousness of the app. © 2012 Massachusetts Institute of Technology.

App developers are constantly competing against each other to win more downloads for their apps. With hundreds of thousands of apps in these online stores, what strategy should a developer use to be successful? Should they innovate, make many similar apps, optimise their own apps or just copy the apps of others? Looking more deeply, how does a complex app ecosystem perform when developers choose to use different strategies? This paper investigates these questions using AppEco, the first Artificial Life model of mobile application ecosystems. In AppEco, developer agents build and upload apps to the app store; user agents browse the store and download the apps. A distinguishing feature of AppEco is the explicit modelling of apps as artefacts. In this work we use AppEco to simulate Apple's iOS app ecosystem and investigate common developer strategies, evaluating them in terms of downloads received, app diversity, and adoption rate.

Requirements elicitation is the software engineering activity in which stakeholder needs are understood. It involves identifying and prioritizing requirements—a process difficult to scale to large software projects with many stakeholders. This paper proposes StakeRare, a novel method that uses social networks and collaborative filtering to identify and prioritize requirements in large software projects. StakeRare identifies stakeholders and asks them to recommend other stakeholders and stakeholder roles, builds a social network with stakeholders as nodes and their recommendations as links, and prioritizes stakeholders using a variety of social network measures to determine their project influence. It then asks the stakeholders to rate an initial list of requirements, recommends other relevant requirements to them using collaborative filtering, and prioritizes their requirements using their ratings weighted by their project influence. StakeRare was evaluated by applying it to a software project for a 30,000-user system, and a substantial empirical study of requirements elicitation was conducted. Using the data collected from surveying and interviewing 87 stakeholders, the study demonstrated that StakeRare predicts stakeholder needs accurately and arrives at a more complete and accurately prioritized list of requirements compared to the existing method used in the project, taking only a fraction of the time.

Software projects often fail because stakeholder communication and involvement are inadequate. This paper proposes a novel method to understand project social networks and their corresponding stakeholder involvement. The method uses five types of model social network, which represent various types of stakeholder activity in a project. It exploits evolutionary computation to correlate the social network of a real software project against each model. Experiments show that the real project most resembles the "rational" model where stakeholders who are more highly connected in the social network are more involved in the project.

Software projects typically rely on system analysts to conduct requirements elicitation, an approach potentially costly for large projects with many stakeholders and requirements. This paper describes StakeSource2.0, a web-based tool that uses social networks and collaborative filtering, a "crowdsourcing" approach, to identify and prioritise stakeholders and their requirements.

Many software projects fail because they overlook stakeholders or involve the wrong representatives of significant groups. Unfortunately, existing methods in stakeholder analysis are likely to omit stakeholders, and consider all stakeholders as equally influential. To identify and prioritise stakeholders, we have developed StakeNet, which consists of three main steps: identify stakeholders and ask them to recommend other stakeholders and stakeholder roles, build a social network whose nodes are stakeholders and links are recommendations, and prioritise stakeholders using a variety of social network measures. To evaluate StakeNet, we conducted one of the first empirical studies of requirements stakeholders on a software project for a 30,000-user system. Using the data collected from surveying and interviewing 68 stakeholders, we show that StakeNet identifies stakeholders and their roles with high recall, and accurately prioritises them. StakeNet uncovers a critical stakeholder role overlooked in the project, whose omission significantly impacted project success.

Projects often fail because they overlook stakeholders. Unfortunately, existing stakeholder analysis tools only capture stakeholders' information, relying on experts to manually identify them. StakeSource is a web-based tool that automates stakeholder analysis. It "crowdsources" the stakeholders themselves for recommendations about other stakeholders and aggregates their answers using social network analysis.

Service-oriented information systems should be built with the understanding of how they meet stakeholder intentions and support enterprise motivation. Existing requirements engineering approaches for services do not address the alignment of service-oriented information systems with business needs. This paper presents a model to support alignment based on the relationships between stakeholder intentions and enterprise motivation. A case study of the Japan Post procuring customer relationship management services illustrates the model.