Christy Lazicky MPA/ID 2017 demonstrates how machine learning can have human results.

By Ramie Jacobson
July 17, 2017


The culmination of the MPA/ID curriculum is the Second Year Policy Analysis (SYPA), a capstone project which offers students the opportunity to deploy the skills acquired during the program, integrate their course work, and provide specific policy recommendations in the context of a concrete development problem. For Christy Lazicky (MPA/ID 2017), the SYPA posed a unique challenge: having taken more data science classes than any other student, how could she apply her prowess for machine learning to her passion for international development? The result of her hard work was an Outstanding SYPA Award-winning paper, “Improving Conflict Early Warning Systems for United Nations Peacekeeping.”

Machine learning—developing computer programs to use data to learn for themselves—is not covered in standard econometrics courses but is especially relevant to economics and social sciences. Having worked in the impact evaluation field at Innovations for Poverty Action (IPA) and the Abdul Latif Jameel Poverty Action Lab (J-PAL) prior to coming to the Kennedy School, Christy took the data science course sequence as a way to fuse her passions for international development and data analysis. By applying her technical toolkit to conflict prediction, Christy was able to write her SYPA on a topic that inspired her while preparing for her career after graduation. 

Christy explained her use of data science in her paper saying, “one application of machine learning methods for prediction to development is predicting conflict outbreak. Currently, many peacekeeping missions rely on a ‘conflict early warning system’ to help them predict future outbreak or escalation of conflict in order to mobilize resources ahead of time to protect civilians. Often these systems are not data-driven, relying on a few human analysts to distill a substantial number of qualitative reports into a conflict risk assessment in a very limited time window. 

Incorporating machine learning methods into early warning systems could overcome these drawbacks by providing a systematic and efficient process for delivering conflict predictions across an entire country. For example, a model could be developed that outputs a conflict risk level (as defined by number of conflict events or fatalities that have occurred in a given area) using data on socioeconomic or geographic features of the area. Examples of such socioeconomic and geographic features that have been shown to be predictive of conflict risk include income level, food security, precipitation and terrain type. 

Using machine learning for predictive analytics in peacekeeping has the potential to help peacekeeping forces more quickly and effectively make predictions about conflict, and thus deploy resources more rapidly to save lives.” 

Christy Lazicky MPA/ID 2017

Classmates and faculty alike were impressed by Christy’s work. One student remarked, “[predicting conflict using data science] is a very important topic in development but one that often falls outside the bounds of the problems to which we apply the MPA/ID skill set.” 

Her section leader, Professor Rema Hanna, wrote, “Ms. Lazicky identified cutting edge models that could be used by the UN in developing a more data-driven early warning system for conflict in Africa. She then compared and assessed the results of the different methodologies she identified against each other and the status quo system. Her political and administrative analysis helped result in a final recommendation that is really actionable for the UN, and which could feasibly be rolled out and piloted. It was truly an impressive piece of work.”

Having graduated from the MPA/ID Program, Christy will begin working as a Project and Technical Manager on with IDinsight in Nairobi, Kenya where she will continue to apply her predictive analytics to the field of development. From wrangling data from administrative sources, cleaning primary collected data, strategically thinking through model selection and which variables to include, training and testing the models, and concisely and compellingly communicating results to clients, Christy will be putting the skills she learned in these courses to good use. 

Originally from New York, Christy received a BA in Mathematics from Dartmouth College and an MPA/ID from the Harvard Kennedy School in 2017.