By Nick Gotts
Jill Lepore (2020) If Then: How One Data Company Invented the Future. John Murray. ISBN: 978-1-529-38617-2 (2021 pbk edition). [Link to book]
This is a most frustrating book. The company referred to in the subtitle is the Simulmatics Corporation, which collected and analysed data on public attitudes for politicians, retailers and the US Department of Defence between 1959 and 1970. Lepore says it carried out “simulation”, but is never very clear about what “simulation” meant to the founders of Simulmatics, what algorithms were involved, or how these algorithms used data. The history of Simulmatics is narrated along with that of US politics and the Vietnam War during its period of operation; the company worked for John Kennedy’s presidential campaign in 1960, although the campaign was shy about admitting this. There is much of interest in this historical context, but the book is marred by the apparent limitations of Lepore’s technical knowledge, her prejudices against the social and behavioural sciences (and in particular the use of computers within them), and irritating “tics” such as the frequent repetition of “If/Then”. There are copious notes, and an index, but no bibliography.
Lepore insists that human behaviour is not predictable, whereas both everyday observation and the academic study of human sciences and history show that on both individual and collective levels it is partially predictable – if it were not, social life would be impossible – and partially unpredictable; she also claims that there is a general repudiation of the importance of history among social and behavioural scientists and in “Silicon Valley”, and seems unaware that many historians and other humanities researchers use mathematics and even computers in their work.
Information about Simulmatics’ uses of computers is in fact available from contemporary documents which its researchers published. In the case of Kennedy’s presidential campaign (de Sola Pool and Abelson 1961, de Sola Pool 1963), the “simulation” involved was the construction of synthetic populations in order to amalgamate polling data from past (1952, 1954, 1956, 1958) American election campaigns. Americans were divided into 480 demographically defined “voter types” (e.g. “Eastern, metropolitan, lower-income, white, Catholic, female Democrats”), and the favourable/unfavourable/neither polling responses of members of these types to 52 specific “issues” (examples given include civil rights, anti-Communism, anti-Catholicism, foreign aid) were tabulated. Attempts were then made to “simulate” 32 of the USA’s 50 states by calculating the proportions of the 480 types in those states and assuming the frequency of responses within a voter type would be the same across states. This produced a ranking of how well Kennedy could be expected to do across these states, which matched the final results quite well. On top of this work an attempt was made to assess the impact of Kennedy’s Catholicism if it became an important issue in the election, but this required additional assumptions on how members of nine groups cross-classified by political and religious allegiance would respond. It is not clear that Kennedy’s campaign actually made any use of Simulmatics’ work, and there is no sense in which political dynamics were simulated. By contrast, in later Simulmatics work not dealt with by Lepore, on local referendum campaigns about water fluoridation (Abelson and Bernstein 1963), an approach very similar to current work in agent-based modelling was adopted. Agents based on the anonymised survey responses of individuals both responded to external messaging, and interacted with each other, to produce a dynamically simulated referendum campaign. It is unclear why Lepore does not cover this very interesting work. She does cover Simulmatics’ involvement in the Vietnam War, where their staff interviewed Vietnamese civilians and supposed “defectors” from the National Liberation Front of South Vietnam (“Viet Cong”) – who may in fact simply have gone back to their insurgent activity afterwards; but this work does not appear to have used computers for anything more than data storage.
In its work on American national elections (which continued through 1964) Simulmatics appears to have wildly over-promised given the data that it would have had available, subsequently under-performed, and failed as a company as a result; from this, indeed, today’s social simulators might take warning. Its leaders started out as “liberals” in American terms, but appear to have retained the colonialist mentality generally accompanying this self-identification, and fell into and contributed to the delusions of American involvement in the Vietnam War – although it is doubtful whether the history of this involvement would have been significantly different if the company had never existed. The fact that Simulmatics was largely forgotten, as Lepore recounts, hints that it was not, in fact, particularly influential, although interesting as the venue of early attempts at data analytics of the kind which may indeed now threaten what there is of democracy under capitalism (by enabling the “microtargeting” of specific lies to specific portions of the electorate), and at agent-based simulation of political dynamics. From a personal point of view, I am grateful to Lepore for drawing my attention to contemporary papers which contain far more useful information than her book about the early use of computers in the social sciences.
Abelson, R.P. and Bernstein, A. (1963) A Computer Simulation Model of Community Referendum Controversies. The Public Opinion Quarterly Vol. 27, No. 1 (Spring, 1963), pp. 93-122. Stable URL http://www.jstor.com/stable/2747294.
de Sola Pool, I. (1963) AUTOMATION: New Tool For Decision Makers. Challenge Vol. 11, No. 6 (MARCH 1963), pp. 26-27. Stable URL https://www.jstor.org/stable/40718664.
de Sola Pool, I. and Abelson, R.P. (1961) The Simulmatics Project. The Public Opinion Quarterly, Vol. 25, No. 2 (Summer, 1961), pp. 167-183. Stable URL https://www.jstor.org/stable/2746702.
Gotts, N. (2023) Yes, but what did they actually do? Review of: Jill Lepore (2020) "If Then: How One Data Company Invented the Future". Review of Artificial Societies and Social Simulation, 9 Mar 2023. https://rofasss.org/2023/03/09/ReviewofJillLepore
© The authors under the Creative Commons’ Attribution-NoDerivs (CC BY-ND) Licence (v4.0)