Micah Altman

Background

Micah Altman was born on August 31, 1967 in St. Louis, Missouri, United States.

Altman studied Computer Science and Political Philosophy at the same time in Brown University, a private Ivy League research university located in Providence, Rhode Island. He graduated magna cum laude in both of his B.A. degrees in 1989. He went to graduate school at California Institute of Technology where he studied Social Science under Morgan Kousser and received a Ph.D. in 1998. Following that, he worked as a postdoctoral researcher in Gary King's research group at Harvard.

In Summer of 1989, Altman started working at Sun Microsystems in Mountain View, California and worked as a Software Engineer. At Sun Microsystems, he developed a Windows systems components called NeWS using an object oriented postscript. He also worked with Silicon Graphics, Inc. as a technical staff from 1990 until 1992. At Silicon, he developed courses, benchmarks, course-ware, on-site consulting, performance computing, parallelization, real time software, and systems and network performance tuning. From 1993 until 1996, he worked as a high performance computing consultant at A-Z Technology.

From 1997 to 2012, he worked at Harvard University, managing administrative positions and conducting research. He was appointed Associate Director of Harvard - MIT Data Center, Archival Director of the Henry A. Murray archive, and Director of data archiving and acquisition of Institute for Quantitative Social Science at Harvard University. He was also awarded a Research fellowship position at the Center for Basic Research in Social Science and worked as a Scientist at the Institute for Quantitative Social Science.

In March 2012, Altman was appointed as Director of Research at the Massachusetts Institute of Technology Libraries and Head Scientist for the Program for Information Science, and as a non-resident senior fellow at the Brookings Institution in Washington, DC.

Altman is currently living in Cambridge, Massachusetts, and has two children.

Electoral districting and redistricting

Altman's contributions to electoral districting and redistricting have been both theoretical and implementational. He studied fundamental aspects of automatability of redistricting for his doctoral research at Caltech and showed that because the number of different ways to partition a region into electoral districts is prohibitively large for all but trivial cases, the computational complexity of the districting problem is NP-hard and hence likely to be intractable without further constraints and heuristics. The proofs are by reduction of the redistricting sub-problems of creating equi-populated districts, maximally compact districts, maximally competitive districts and contiguous equal-population districts, to the problems of 3-partitioning, distance-d partition of points in the plane, minimum sum of squares, and the cut into connected components of bounded weight, respectively.

The undesirable implications of this result are that redistricting cannot be fully automated in practice and the choice of constraints and manual selection of the winning, "optimal" plan from a group of auto-generated plans, reintroduce value-laden and politically biased decision making back into the redistricting process (something that the use of "objective" computer programs was hoped to avoid), while potentially also legitimizing such undercover gerrymandering for the less knowledgeable public.

The computational simulations that he performed showed also that even the constraints that have been traditionally considered politically non-preferential, such as the overall compactness of the district, are not necessarily non-preferential because compactness requirements have different effects on political groups if the groups are distributed in geographically different ways. This result was referenced by the Supreme Court justices in the Vieth v. Jubelirer case.

Altman and his colleagues later created the open-source BARD software and the DistrictBuilder software to enable users to automatically determine district boundaries on the basis of demographic data (voting age, race, medium income, education) and other criteria such as district compactness and contiguity. They address the computational complexity of the districting problem by using metaheuristics (such as simulated annealing, genetic algorithms, tabu search and greedy randomized adaptive search) for refining auto-generated or pre-existing plans, and implement different performance enhancements like evaluation caching, explicit memory management, and distributed computing. These programs minimize but do not eliminate the necessary human intervention in narrowing down the number of district plans.

Data curation

Altman's research in data curation has been in relation to his work at Harvard libraries and data archives, especially the Virtual Data Center project that he led with Sidney Verba, and with its successor - the Dataverse Network. He has studied ways of improving the methodologies for preserving, archiving and cataloguing research data in social sciences, and methods for distributing and disseminating data for reuse by other researchers. To yield reliable and comparable results, standard methods of data encoding are needed for data attribution and data citation, and for maximally accurate data verification and replication.

In "A Proposed Standard for the Scholarly Citation of Quantitative Data" by Altman and King in 2007, the authors proposed a standard for citing quantitative data, similar to the existing standards for citing papers and analyses that are performed on the data, as no such standard for data citation existed before. The citation information they recommended included a unique global identifier, a short character string guaranteed to be unique among all such identifiers, that permanently identifies the data set independent of its location, and a universal numeric fingerprint, a fixed-length string of numbers and characters that summarize all the content in the data set, such that a change in any part of the data would produce a different fingerprint.

The data fingerprints they propose are based on checksums and can be created by applying hash functions to normalized and approximated data, and used in statistical applications to prevent misinterpretations of data, and to verify content and format during data migration and archiving. The algorithm for generating the fingerprints has undergone several revisions because the initial versions underestimated the expressive power needed to encode the data and the simpler algorithm inherited the weaknesses of the MD5 hash function that was shown to have several vulnerabilities.

Statistical computing

Since there are a great number of variables involved in experiments in social sciences and the values of these variables are often entangled, complex or hard to quantify, precise predictions are hard to make. In "Numerical Issues in Statistical Computing for the Social Scientist" by Altman, Gill, and McDonald in 2004, an advanced-level reference book for social scientists about computational statistics, shows that these problems are frequently compounded by measurement errors and numerical inaccuracies that arise in statistical computing. The sources of these errors range from un-modeled measurement errors to software errors in statistical packages, errors in data input, data that is ill-conditioned for a particular model, floating point underflows and overflows, rounding errors, non-random structures in random number generators, local optima or discontinuities in optimization, inappropriate or unlucky choices of starting values and inadequate stopping criteria.

It is shown that the knowledge of numerical methods underlying computerized statistical calculations and how they are used in statistical packages is essential for planning quantitative studies in social sciences and for making accurate inferences, and techniques and diagnostic tests are offered to detect such problems and prescriptions for good statistical computing practice that results in greater accuracy, precision, robustness, sensitivity and reproducibility.

Awards and recognitions

Altman received the "Weaver Award" for the best paper in representation and electoral systems in 1998. The award was given by the American Political Science Association, a professional association of political science students and scholars in the United States. In 1999, he received the "Outstanding Dissertation Award" from the Western Political Science Association for his doctoral dissertation, "Districting Principles and Democratic Representation". In the same year, he also received the "Best Professional Political Science Website" from the American Political Science Association along with his other website teams. In 2005 and 2009, he received the "Best Research Software" award for his The Virtual Data Center software, and "Best Instructional Software" for his Better Automated Redistricting software respectively. Both awards were given by the American Political Science Association. In 2011, he was recognized with the "Best Policy Innovations of 2011" by Politico for his Public Mapping Project called the DistrictBuilder. He created the software with Michael McDonald. In 2012, he received the "Outstanding Software Development" award from American Political Science Association. Also in the same year, he received the "Data Innovation Award for Social Impact" from the O’Reilly Strata Conference. Most recently, Altman was awarded the 2013 Antonio Pizzigati Prize for Software in the Public Interest for his work developing software that encourages transparency and public participation in the electoral redistricting process.

Altman works have also been cited by the United States Supreme Court, and have been covered by multiple local and national media organizations.

Selected works

1. Micah Altman (1997). Is Automation the Answer: The Computational Complexity of Automated Redistricting. Rutgers Computer and Technology Law Journal. pp. 81–142. 
2. Micah Altman (1998). Districting Principles and Democratic Representation. Pasadena, California, United States: California Institute of Technology. p. 367. 
3. Micah Altman; et al. (2004). Numerical Issues in Statistical Computing for the Social Scientist. John Wiley & Sons. ISBN 0-471-23633-0. 

1. Micah Altman (November 1998). "Modeling the Effect of Mandatory District Compactness on Partisan Gerrymanders". Political Geography 17 (8): 989–1012. 
2. Micah Altman & Michaell P. McDonald (September 2001). "Choosing Reliable Statistical Software". Political Science & Politics 34 (3): 681–687. 
3. Micah Altman; et al. (August 2005). "From Crayons to Computers". Social Science Computer Review 23 (3): 334–346. 
4. Micah Altman & Gary King (March 2007). "A Proposed Standard for the Scholarly Citation of Quantitative Data". D-Lib Magazine 13 (3 & 4): 1082–9873. 
5. Micah Altman (2008). "A Fingerprint Method for Scientific Data Verification". Advances in Computer and Information Sciences and Engineering: 311–316. 
6. Micah Altman & Michael P. McDonald (2011). "BARD: Better Automated Redistricting". Journal of Statistical Software.