Tuesday 22nd November 2016 - Pre-Conference Workshops
Workshop 1: Quantifying uncertainty with structured expert judgement
An SRA ANZ workshop with Anca. M. Hanea
Expert elicitation is a broad field which includes asking experts for specific information (facts, data, sources, requirements, etc.) or for expert judgements about things (preferences, utilities, probabilities, estimates, etc). This workshop covers some aspects (detailed below) of the elicitation and combination of expert judgements about uncertain events or quantities.
In this context, a variety of judgements could be elicited. An obvious judgement is an estimate. However, simply eliciting an estimate gives no indication of the expert‘s degree of uncertainty or confidence regarding this estimate.
There are strong theoretical and practical arguments to say that the proper representation of an expert‘s knowledge about an uncertain quantity is a probability distribution. A probability distribution quantifies the expert‘s knowledge and uncertainty in detail.
Challenges arise when judgements are elicited from multiple experts and the model requires a single probability distribution, which means that the various judgements must be aggregated. This aggregation can be done by the experts themselves, through a process of interaction that is designed to encourage the experts to reach a consensus view (behavioural aggregation). Alternatively, it may be done externally, by applying an aggregation formula (mathematical aggregation). A third option is to use a combination of the two approaches mentioned above.
One of the most popular mathematical aggregation schemes is a weighted linear combination of judgements. Choosing the weights of this combination is a critical step of the mathematical aggregation. Maybe the most popular weighting scheme uses equal weights, whereas a slightly more complex method of defining weights is based on the experts’ relative performance on a set of test variables.
Expert elicitation is a broad field which includes asking experts for specific information (facts, data, sources, requirements, etc.) or for expert judgements about things (preferences, utilities, probabilities, estimates, etc). This workshop covers some aspects (detailed below) of the elicitation and combination of expert judgements about uncertain events or quantities.
In this context, a variety of judgements could be elicited. An obvious judgement is an estimate. However, simply eliciting an estimate gives no indication of the expert‘s degree of uncertainty or confidence regarding this estimate.
There are strong theoretical and practical arguments to say that the proper representation of an expert‘s knowledge about an uncertain quantity is a probability distribution. A probability distribution quantifies the expert‘s knowledge and uncertainty in detail.
Challenges arise when judgements are elicited from multiple experts and the model requires a single probability distribution, which means that the various judgements must be aggregated. This aggregation can be done by the experts themselves, through a process of interaction that is designed to encourage the experts to reach a consensus view (behavioural aggregation). Alternatively, it may be done externally, by applying an aggregation formula (mathematical aggregation). A third option is to use a combination of the two approaches mentioned above.
One of the most popular mathematical aggregation schemes is a weighted linear combination of judgements. Choosing the weights of this combination is a critical step of the mathematical aggregation. Maybe the most popular weighting scheme uses equal weights, whereas a slightly more complex method of defining weights is based on the experts’ relative performance on a set of test variables.
Workshop 2: Fatigue Risk Management
An SRA ANZ workshop with Professor Sally Ferguson and Associate Professor Matthew Thomas
Fatigue is defined as performance impairment that results from inadequate sleep, disruption to our body clock, or excessive time on task. Traditionally, fatigue has been managed in the work environment by prescriptive rules that limit the number of hours that can be worked, and specifying minimum rest periods between work. However, recently this approach has been demonstrated to not always be effective in managing the risks associated with fatigue to a level that is as low as reasonably practical. This workshop will explore current best-practice in fatigue risk management. Opening with an overview of the science of human sleep and fatigue, the workshop will then explore methods used across a wide range of industries to identify risks associated with fatigue and the ways in which these risks can be effectively managed through a range of risk mitigation strategies. The workshop will provide participants with a solid foundation in the science of fatigue, and novel methods used to assess and control fatigue-related risk.
Fatigue is defined as performance impairment that results from inadequate sleep, disruption to our body clock, or excessive time on task. Traditionally, fatigue has been managed in the work environment by prescriptive rules that limit the number of hours that can be worked, and specifying minimum rest periods between work. However, recently this approach has been demonstrated to not always be effective in managing the risks associated with fatigue to a level that is as low as reasonably practical. This workshop will explore current best-practice in fatigue risk management. Opening with an overview of the science of human sleep and fatigue, the workshop will then explore methods used across a wide range of industries to identify risks associated with fatigue and the ways in which these risks can be effectively managed through a range of risk mitigation strategies. The workshop will provide participants with a solid foundation in the science of fatigue, and novel methods used to assess and control fatigue-related risk.