Hybrid attacks coordinate the exploitation of vulnerabilities across domains to undermine trust in authorities and cause social unrest. Whilst such attacks have primarily been seen in active conflict zones, there is growing concern about the potential harm that can be caused by hybrid attacks more generally and a desire to discover how better to identify and react to them. In addressing such threats, it is important to be able to identify and understand an adversary’s behaviour. Game theory is the approach predominantly used in security and defence literature for this purpose. However, the underlying rationality assumption, the equilibrium concept of game theory, as well as the need to make simplifying assumptions can limit its use in the study of emerging threats. To study hybrid threats, we present a novel agent-based model in which, for the first time, agents use reinforcement learning to inform their decisions. This model allows us to investigate the behavioural strategies of threat agents with hybrid attack capabilities as well as their broader impact on the behaviours and opinions of other agents.

Model of influence of access to social information spread via social network on decisions in a two-person game.

Package for simulating the behavior of experts in a scientific-forecasting competition, where the outcome of experiments itself depends on expert consensus. We pay special attention to the interplay between expert bias and trust in the reward algorithm. The package allows the user to reproduce results presented in arXiv:2305.04814, as well as testing of other different scenarios.

Machine learning technologies have changed the paradigm of knowledge discovery in organizations and transformed traditional organizational learning to human-machine hybrid intelligent organizational learning. However, it remains unclear how human-machine trust, which is an important factor that influences human-machine knowledge exchange, affects the effectiveness of human-machine hybrid intelligent organizational learning. To explore this issue, we used multi-agent simulation to construct a knowledge learning model of a human-machine hybrid intelligent organization with human-machine trust.

This model has been created with and for the researcher-farmers of the Muonde Trust (http://www.muonde.org/), a registered Zimbabwean non-governmental organization dedicated to fostering indigenous innovation. Model behaviors and parameters (mashandiro nemisiyano nedzimwe model) derive from a combination of literature review and the collected datasets from Muonde’s long-term (over 30 years) community-based research. The goals of this model are three-fold (muzvikamu zvitatu):
A) To represent three components of a Zimbabwean agro-pastoral system (crops, woodland grazing area, and livestock) along with their key interactions and feedbacks and some of the human management decisions that may affect these components and their interactions.
B) To assess how climate variation (implemented in several different ways) and human management may affect the sustainability of the system as measured by the continued provisioning of crops, livestock, and woodland grazing area.
C) To provide a discussion tool for the community and local leaders to explore different management strategies for the agro-pastoral system (hwaro/nzira yekudyidzana kwavanhu, zvipfuo nezvirimwa), particularly in the face of climate change.

An agent model is presented that aims to capture the impact of cheap talk on collective action in a commons dilemma. The commons dilemma is represented as a spatially explicit renewable resource. Agent’s trust in others impacts the speed and harvesting rate, and trust is impacted by observed harvesting behavior and cheap talk. We calibrated the model using experimental data (DeCaro et al. 2021). The best fit to the data consists of a population with a small frequency of altruistic and selfish agents, and mostly conditional cooperative agents sensitive to inequality and cheap talk. This calibrated model provides an empirical test of the behavioral theory of collective action of Elinor Ostrom and Humanistic Rational Choice Theory.

This model examines how financial and social top-down interventions interplay with the internal self-organizing dynamics of a fishing community. The aim is to transform from hierarchical fishbuyer-fisher relationship into fishing cooperatives.

This is a simulation model of communication between two groups of managers in the course of project implementation. The “world” of the model is a space of interaction between project participants, each of which belongs either to a group of work performers or to a group of customers. Information about the progress of the project is publicly available and represents the deviation Earned value (EV) from the planned project value (cost baseline).
The key elements of the model are 1) persons belonging to a group of customers or performers, 2) agents that are communication acts. The life cycle of persons is equal to the time of the simulation experiment, the life cycle of the communication act is 3 periods of model time (for the convenience of visualizing behavior during the experiment). The communication act occurs at a specific point in the model space, the coordinates of which are realized as random variables. During the experiment, persons randomly move in the model space. The communication act involves persons belonging to a group of customers and a group of performers, remote from the place of the communication act at a distance not exceeding the value of the communication radius (MaxCommRadius), while at least one representative from each of the groups must participate in the communication act. If none are found, the communication act is not carried out. The number of potential communication acts per unit of model time is a parameter of the model (CommPerTick).

The managerial sense of the feedback is the stimulating effect of the positive value of the accumulated communication complexity (positive background of the project implementation) on the productivity of the performers. Provided there is favorable communication (“trust”, “mutual understanding”) between the customer and the contractor, it is more likely that project operations will be performed with less lag behind the plan or ahead of it.
The behavior of agents in the world of the model (change of coordinates, visualization of agents’ belonging to a specific communicative act at a given time, etc.) is not informative. Content data are obtained in the form of time series of accumulated communicative complexity, the deviation of the earned value from the planned value, average indicators characterizing communication - the total number of communicative acts and the average number of their participants, etc. These data are displayed on graphs during the simulation experiment.
The control elements of the model allow seven independent values to be varied, which, even with a minimum number of varied values (three: minimum, maximum, optimum), gives 3^7 = 2187 different variants of initial conditions. In this case, the statistical processing of the results requires repeated calculation of the model indicators for each grid node. Thus, the set of varied parameters and the range of their variation is determined by the logic of a particular study and represents a significant narrowing of the full set of initial conditions for which the model allows simulation experiments.
…

This model implements a Bayesian belief revision model that contrasts an ideal agent in possesion of true likelihoods, an agent using a fixed estimate of trusting its source of information, and an agent updating its trust estimate.

The network-based trust game is a hybridization of both the repeated trust games and the network games.