C2-g Services

Provide an innovative data pipeline to improve decisions

banner

1. How do we quantify human performance?

Perception-action cycles (PACs) provide a robust and succinct representation of human performance across several operational settings. If represented correctly, causal (e.g., Bayesian) networks can quantitatively capture human PACs, and offer critical insight into how humans convert information to decisions.

How do we obtain appropriate data to train these models?

Many times, an adequate amount real-world data are not available to train causal models of human performance. Therefore, we must sometimes leverage an inexpensive approach for eliciting the data required for developing our models. C2-g has vast experience in designing low-fidelity methods (e.g., computational games) to produce data necessary for training causal models.

2. Do these models predict real-world performance?

Since the models were trained using low-fidelity data, it is possible that they won't predict operational performance due to differences between the real-world and simulated environment, or because of personal characteristics between actual operators and low-fidelity participants (e.g., training).

How do we overcome these challenges?

Multifidelity methods offer a solution, as they leverage both low- and high-fidelity data to maximize the accuracy of model estimates, while minimizing the cost associated with parametrization. They quantitatively adjust model parameters to leverage the strengths of each data source, while attempting to overcome the limitations.

3. How can we use the combination of these approaches?

This innovative pipeline can be used to design decision-support systems that are safe, adaptive and actionable. More specifically, it provides insight into the type and amount of information to present people, given their operational context. It does this by learning models of how people encode and fuse information to form beliefs, and how that information needs to be updated across context. Moreover, it does NOT use operator preferences to select information, as that can result in perpetuating or even amplifying biases.

Any other ways this can be used?

Since we are leveraging simulation data to train our low-fidelity models, we can begin to evaluate the risk associated with introducing new technology into the operational setting at the concept stage of development. Normally, these data aren't available until after the technology has been deployed, so this saves a great deal of time and money and assures only the safest and most effective technology is developed.

4. What are other services you offer?

We also provide general data science consulting services to academics, businesses and government. We are especially keen to assist with charity organizations by offering suggestions about how they can collect and use data to improve their operations to maximize the impact of their services. We are also happy to set-up analysis scripts for your company and train key employees how to use and edit the files. We have a great deal of experience teaching complex topics in an approachable and intuitive manner.

We love our privacy, why should we share our data with an outsider?

C2-g operates under the most rigorous ethical standards in the industry. We will be happy to sign and honor any NDAs that are needed by your company, and would never disclose trade secrets to people outside your company.