When Louis XIV saw the new maps of France he sponsored in 1693 he supposedly complained that his cartographers had cost his kingdom more land in a year than foreign armies had done in a century.
(From Paul Slack’s The Invention of Improvement)
–Long-held heuristics, like routines and standard operating procedures, are shorthand ways of doing things without all the uncertainty of reinventing the wheel. Newer heuristics include big data algorithms we don’t understand and policy narratives we think we do, both of which enable making decisions in the face of uncertainty. Both shorthands are treated as good-enough, like a new atlas of maps.
The older heuristics are relied upon because they are said to reduce uncertainty; more recent ones are used to better manage (in the face of) uncertainty that hasn’t been/cannot be reduced, at least for the moment.
One major commonality between both the old and new needs to be highlighted. While typically not taken as such, both are less a shorthand than clues for what to do ahead.
–By way of example, handbooks detailing how to respond to unpredictable floods and famines were written by and for administrators in Imperial China. Over the course of some thousand years, handbooks started to group together what had been learned into tables of maxims (sometime cast in rhyme) for ease of reference by users.
Handbooks “are quick to insist, however, that using the tables is not sufficient in the long run: for the professional administrator they are rather a ‘clue’. . .that indicates where to go in the more complete texts,” writes Pierre-Étienne Will, the most recent and comprehensive bibliographer of handbooks (2020, XLIV). This status of heuristic-as-clue is to alert us to important omissions that require reference beyond any shorthand exposition (Ibid, 568). Occasions when a map proves imperfect and misleading are all too familiar.
Professor Will elaborates in an email: “’clue’ (yinxian 引線. . .) literally means ‘a thread that leads to…’, ‘that can be pulled to get…, or something of the sort. The same character yin is part of the words suoyin and yinde, meaning ‘index’, in modern Chinese. The tables or rhymes are like indexes to the complete texts.”
–I want to apply this notion of heuristic-as-clue more speculatively to the newer algorithms derived from big data. We’re told that, even though the algorithms are not based on models of known cause and effect, they identify complex, albeit opaque, correlations said to be worth relying upon.
But that stops short of the needful. The status of a heuristic as clue underscores that, just as with causal models, there’s also a great deal yet to puzzle out with correlations before going forward. Correlations are not just the start of an analysis. They also are in context and those contexts start the analysis as well. Correlations index spatial-temporal references to be pursued.
–There is no obvious point of entry when it comes to revealing the wider references. For purposes of illustration, start with the canonical index of fire, smoke. In the same way, the heat from server centers (some indeed call it “data exhaust”) indexes the large electricity usage in generating and updating algorithms. But context doesn’t stop there. Other clues are less spatial-temporal and more social for the heuristic inseparable from wider referential meanings.
Again, by way of example, the status of the algorithm-as-heuristic clues you into the underlying assumptions for using big dataset algorithms, not least of exemplify “trust.” Some say, e.g.:
- algorithms deliver the best result among the other methods and heuristics available;
- while not free of bias, they do a better job than others by virtue of the huge run of cases and calculations;
- some kind of result at the scale of big data is better than no result, plus the algorithmic result is often more timely; and
- anyway, there’s always a danger that the critics of big data algorithms take them more seriously than the users, like consumers who comparison-shop and then make their own decisions.
The wider point here is that the methodological duty of care in using heuristics means treating them as indexes of that which cannot be omitted, yet could have already been omitted, from analysis and practice when usefulness is the question.
Pierre-Étienne Will (2020). “Introduction,” in: Handbooks and Anthologies for Officials in Imperial China: A descriptive and critical bibliography. Koninklijke Brill NV, Leiden, The Netherlands.
The notion of index in the sense of smoke and fire follows that of C.S. Pierce (“purse”), a founder of American pragmatism.
R. Machen and E. Nost (2021). “Thinking algorithmically: The making of hegemonic knowledge in climate governance.” Transactions of the Institute of British Geographers: 1-15.
Julia Velkova (2021). “Thermopolitics of data: Cloud infrastructures and energy futures.” Cultural Studies.