CASOS Members, ISR, CASOS IGERT
Email: hirshman[AT]cs.cmu.edu
CMU Affiliation: CASOS
Office: Wean Hall 5119
Phone: 412-268-4682
Thesis Abstract: While social simulation models are more powerful than they were sixty years ago, modelers continue to face a persistent problem: what should we do when we lack empirical data for a given parameter? Current computational models are bigger, faster, more detailed, more realistic, and more targeted than ever. Though their theoretical underpinnings have improved, the data that supports them continues to suffer from widely-acknowledged problems. Despite the improvements in simulation logic, the adage “garbage in, garbage out” continues to be relevant.
Simulation designers have traditionally employed one of three strategies for specifying a parameter when no appropriate data is available. Unknown variable values are generated using an appropriate distribution, selected following a sensitivity analysis over the potential values, or specified by a domain or subject matter expert. This thesis suggests that experiment designers also consider another approach: using theory, in conjunction with mathematical techniques, to transform data from a related field into a form appropriate for the model. By employing theory in this way, the modeler can adapt such data using a process that I will call /alignment/. By aligning data from a different domain to specify a variable value, designers can broaden the types of input data they may use if their choices are supported by a strong theoretical foundation.
This thesis seeks to improve inputs to knowledge diffusion models in order to increase overall realism and to understand whether such improvements lead to robust changes in outcome metrics. Specifically, it will investigate the effectiveness of aligning input properties and networks required by a simulation. It will consider the appropriateness of using Iterative Proportional Fitting (IPF), Naïve Bayes (NB), and Non-Negative Matrix Factorization (NMF) to specify certain types of simulation input, and determine whether and how such alignment affect simulation outcome metrics. Sensitivity analysis will assess the effect of staging and simulated population size on the proposed techniques.
Thesis Proposal: Alignment In, Advancement Out: improving inputs to knowledge diffusion simulations
Poster for Thesis Proposal: August 18, 2010, 12:00pm, GHC 7101
Projects:
- Assessing and Changing Population Behavior - Team Member
- Belief Diffusion - Team Member
- Construct - Team Member
Hirshman, Brian & Tang, Jessica A & Jones, Laurie & Proudfoot, James A & Carley, Kathleen M & Marshall, Lawrence & Carter, Bob S & Chen, Clark C . (2016). Impact of Medical Academic Genealogy on Publication Patterns: An analysis of the Literature for Surgical Resection in Brain Tumor Patients. Annals of Neurology, Wiley. [DOI] WebSite: [link]
Hirshman, Brian & St. Charles, Jesse & Carley, Kathleen M . (2011). Leaving us in tiers: can homophily be used to generate tiering effects?. Computational and Mathematical Organization Theory, 17, 318-343. Springer. [DOI] [pdf]
Carley, Kathleen M & Hirshman, Brian. (2011). Agent-Based Models. Encyclopedia of Social Networks. Ed. George A. Barnett. Thousand Oaks, CA: SAGE, 2011. 12-17. SAGE Reference Online., WebSite: [link]
Carley, Kathleen M & Robertson, Dawn & Martin, Michael & Lee, Ju-Sung & St. Charles, Jesse & Hirshman, Brian. (2010). Predicting Intentional and Inadvertent Non-compliance. IRS Research Conference, Washington, DC, June 29-30, 2010., [pdf]
Hirshman, Brian & Morgan, Geoffrey & St. Charles, Jesse & Carley, Kathleen M . (2010). Construct Demo Input Deck. Carnegie Mellon University, School of Computer Science, Institute for Software Research, Technical Report, CMU-ISR-10-118., [pdf]
Carley, Kathleen M & Levis, Alexander H & Moon, Il-Chul & Morgan, Geoffrey & St. Charles, Jesse & Hirshman, Brian & Lanham, Michael & Papantoni-Kazakos, P & Wagenhals, Lee W & Zaidi, Abbas K & Cioffi-Revilla, Claudio & Elder, Rovert J & Levitt, R & AbuSharekh, Ashraf & Kansal, Smriti K & Erkin Olmez, A & Mansoor, Faisal & Faraz Rafi, M & Romano, Pedro & Pham, John. (2010). Computational Modeling of Cultural Dimensions in Adversary Organizations. Fairfax, VA: The Volgenau School of Engineering Dept. of Electrical and Computer Engineering System Architectures Laboratory, George Mason University, Technical Report., WebSite: [link]
Hirshman, Brian & Carley, Kathleen M & Hirshman, Brian. (2009). Variables, Decisions, and Scripting in Construct. Carnegie Mellon University, School of Computer Science, Insitute for Software Research, Technical Report CMU-ISR-09-126, [pdf]
Hirshman, Brian & Carley, Kathleen M & Hirshman, Brian. (2009). Variables, Decisions, and Scripting in Construct. Carnegie Mellon University, School of Computer Science, Insitute for Software Research, Technical Report CMU-ISR-09-126, [pdf]
Carley, Kathleen M & Martin, Michael & Hirshman, Brian. (2009). The Etiology of Social Change. Topics in Cognitive Science, 1, 621-650. [link]
Hirshman, Brian & Martin, Michael & Birukou, Alaiksandr & Bigrigg, Michael & Carley, Kathleen M . (2008). The Impact of Educational Interventions on Real & Stylized Cities. Carnegie Mellon University, School of Computer Science, Institute for Software Research, Technical Report, CMU-ISR-08-114, [pdf]
Hirshman, Brian & Martin, Michael & Carley, Kathleen M . (2008). Modeling Information Access in Construct. Carnegie Mellon University, School of Computer Science, Institute for Software Research, Technical Report CMU-ISR-08-115, [pdf]
Hirshman, Brian & Martin, Michael & Bigrigg, Michael & Carley, Kathleen M . (2008). The Impact of Educational Interventions by Socio-Demographic Attribute. Carnegie Mellon University, School of Computer Science, Institute for Software Research, Technical Report, CMU-ISR-08-118., [pdf]
Hirshman, Brian & Carley, Kathleen M & Kowalchuck, Michael. (2007). Specifying Agents in Construct. Carnegie Mellon University, School of Computer Science, Institute for Software Research, Technical Report CMU-ISRI-07-107., [pdf]
Hirshman, Brian & Carley, Kathleen M & Kowalchuck, Michael. (2007). Loading Networks in Construct. Institute for Software Research, School of Computer Science, Carnegie Mellon University, CASOS Technical Report, [pdf]
