SSC2018 workshops – Monday August 20th, 13.00 – 17.00
1. Tackling Prediction in Empirical Agent-Based Models
Prediction is not the only reason to build a model of any kind, never mind an agent-based one, but it is a very important one. Prediction is also expected in a number of scenarios where models are used with stakeholders, and is a natural thing to consider reasonable in empirical contexts. However, there are a diversity of meanings of “prediction” from a single, quantitative accurate statement of future fact, to qualitative statements of possible futures. As agent-based modellers, we are rightly more circumspect about prediction, as we are more aware of and accustomed to addressing inherent system complexity of the kind that certainly makes the single, quantitative and accurate statement interpretation of prediction unrealistic. This, however, does not stop modellers from alternative traditions, some of whom may consider complexity to be ‘noise’ or simply a matter of coarsening grain, from making predictions based on models we would consider oversimplified and inadequate to the task. If agent-based modelling is to become mainstream, we need to tackle prediction. We need to develop our own language and terminology around prediction, and to understand what predictions with ABMs might look like. The purpose of this workshop is to explore the reluctance of the ABM community to engage with prediction, and to consider ways forward that will address it.
Organizer: Gary Polhill <Gary.Polhill@hutton.ac.uk>
2. Running high performance simulations with Julia programming language
The goal of this workshop is to help social scientists to leverage the power of Julia language to more efficiently carry out large scale agent-based simulation experiments.
We will start with simple examples but in the end we will show how Julia can be used to run huge scale massively parallelized social simulations spreading across many servers in the cloud or on a Cray Supercomputer.
The workshop plan includes:
– introduction to numerical computing and simulations in Julia
– parallelizing simulation models with multithreading
– parallelizing simulation models with multiprocessing
– designing and running simulations on computing clusters in the cloud and supercomputers
The workshop will be hands-on for those who are interested to follow the examples on their computers. The participants who want to participate interactively should:
1. to run computations locally: install Julia language on a computer (https://julialang.org/downloads/)
2. to run computations in the cloud:
2.1. Setup an account on Amazon Web Services cloud computing (https://portal.aws.amazon.com/billing/signup). Please note that in some cases setting up an AWS acount can take few hours (before it gets accepted by AWS team) and hence we advise you to create the AWS account at least one day before the workshop.
2.2. Bring a laptop and have a SSH client installed. On Mac OS and Linux, the SSH client is in-built while on Windows platform the best option is to use SSH that is automatically installed with Git software (https://git-scm.com/download/win). Additionally, Windows users can consider installing ConEmu console that is a much better option than the standard Command Prompt.
Organizers: Przemyslaw Szufel <pszufe@gmail.com>, Bogumił Kamiński <bogumil.kaminski@gmail.com>
3. Social complexity and laboratory experiments – testing assumptions and predictions of social simulation models with experiments
The aims of this workshop are to collect and discuss arguments and examples of research supporting and challenging the usefulness of laboratory experiments and to foster the discussion about strengths and weaknesses of experiments relative to other empirical approaches. To this end, we plan to have a 4 hours workshop in which the morning will be devoted to presentations (maximum 30 minutes each) describing alternative positions about and experiences with experiments testing simulation models. In the second part of the workshop, a roundtable session (with three speakers) will be followed by an open discussion in which competing views will be debated. The aim is to compile a joint review paper about the strengths and weaknesses of laboratory experiments that describes future challenges and opportunities of this empirical method in social simulation research.
Organizer: Francesca Giardini <f.giardini@rug.nl>
4. Using theories on human decision-making in ABMs
The three-hours workshop is organized in two parts. In the first part, we will address the question how to find and select an “appropriate” theory on human decision-making for a given model context, model purpose and research question. This step is closely linked with the questions to which extent qualitative and / or quantitative data can be integrated and to which level of detail human decision-making is modelled. We aim at enabling participants to find a theory by improving language around decision making elements and processes in a model/theory. We will give an input presentation to share our own experiences of using different theories of human decision-making and the impact of theories on model outcomes. This will be followed by a group discussion in which workshop participants can share their experiences and the modelling challenges they face. Ideally, this discussion would be summarized in a decision tree on how to select theories on human decision-making and related data.
In the second part of the workshop, we will move on to the question how to formalize theories on human decision-making and the choices that need to be made once one (or more) theory/theories have been selected. Depending on group size, we will select two to three theories and discuss them in breakout groups.
The breakout groups will collect:
Which criteria are needed for the decision-making? How are they perceived?
How to select a behavioral option? This relates to decision-making based on a single criterion versus multiple criteria.
How to treat time? Here, topics such as scheduling, prediction and time discounting, memory, learning and forgetting are relevant.
We will then compare results of the breakout groups and conclude the workshop. Results of the workshop may feed into a joint publication of interested workshop participants.
Organizer: Nina Schwarz <n.schwarz@utwente.nl>