Template:M intro systems financialisation: Difference between revisions
Jump to navigation
Jump to search
Amwelladmin (talk | contribs) Tags: Mobile edit Mobile web edit Advanced mobile edit |
Amwelladmin (talk | contribs) No edit summary Tags: Mobile edit Mobile web edit Advanced mobile edit |
||
| Line 21: | Line 21: | ||
But we are talking metaphorically here: there is “financialisation” in a broader sense that need not involve ''money'' as such: SAT scores, A-level grades, Out-of-five product reviews, [[performance appraisal]]s, [[RAG status]]es, [[net promoter score]]s, QR codes, implied carbon footprints, gender pay differentials — any numerically measurable criteria that convert the messy, idiosyncratic, intractable ''life experience'' into ordered columns, pivot tables, and scatter plots that can be averaged, extrapolated, enriched, [[Pareto triage]]d, [[standard deviation]] plotted, and put into ranked, tranched ''order''. | But we are talking metaphorically here: there is “financialisation” in a broader sense that need not involve ''money'' as such: SAT scores, A-level grades, Out-of-five product reviews, [[performance appraisal]]s, [[RAG status]]es, [[net promoter score]]s, QR codes, implied carbon footprints, gender pay differentials — any numerically measurable criteria that convert the messy, idiosyncratic, intractable ''life experience'' into ordered columns, pivot tables, and scatter plots that can be averaged, extrapolated, enriched, [[Pareto triage]]d, [[standard deviation]] plotted, and put into ranked, tranched ''order''. | ||
When you are building a technologised process — seeking inputs and calculating outputs — ''[[free text is not your friend]]''. | When you are building a technologised process — seeking inputs and calculating outputs — ''[[free text is not your friend]]''. You can’t ''do'' anything with free text, beyond bucketing it as “other”. “Other” conceals an infinitude of richness and variability. So does “A*”, “Male”, “needs improvement”, “amber” and “British Asian”. | ||
By our process design, we are elect to assign that variability a numerical value of zero within a category, and 1 beyond it, even though the actual variability might transgress the original categorisation. A “male” and a “female” might have more in common than they differ, depending on the reason for categorisation. (For example, when the categorisation is of “financial services professionals” this is almost certainly the case.) | |||
In this way, our own model determines the types of biases we see as much as the data. (We will never know if recreational cricketers, left-handers, introverts, or people who live more than twenty kilometres away are discriminated against because this data is not gathered). | |||
====There are no straight lines in nature==== | ====There are no straight lines in nature==== | ||
{{drop|I|n which we}} call to mind Robin Williams’ great scene in {{br|Dead Poets Society}}, and restate it: just as ''[[there is no machine for judging poetry]]'', there is no machine for judging ''commerce'' either. | {{drop|I|n which we}} call to mind Robin Williams’ great scene in {{br|Dead Poets Society}}, and restate it: just as ''[[there is no machine for judging poetry]]'', there is no machine for judging ''commerce'' either. | ||
Any [[metrics]], balance sheets, [[org chart]]s, projections or discounted cashflow analysis — ''any'' [[formal]] accounting for the intensely human activity of doing business — jettisons much of | Any [[metrics]], balance sheets, [[org chart]]s, projections or discounted cashflow analysis — ''any'' [[formal]] accounting for the intensely human activity of doing business — jettisons much of potential value. The [[map]] can never be more than a brief schematic: it cannot convey the grandeur — or the ''horror'' — of the [[territory]]. The jettisoning is part of the exercise: it is, itself to judge what is and is not important. It is to extract a [[signal]] from [[hubbub|noise]]. That signal is ''ad hoc'', imaginary, a creative work, and by no means exclusive to the hubbub: there are ''infinite'' array of signals we could take out of the hubbub; the ones we do are determined by our cultural fabric, which is made of all the decisions, signals, and institutions we have already built. | ||
This relativity terrifies “right-thinking people”, but there is no way around it: it is best just to ''[[ignore]]'' it: bucket it up with all that other irrelevant and inconvenient stuff, as “other”. | |||
====Map and territory==== | ====Map and territory==== | ||
{{Drop|T|hat much is}} obvious: it is not the lesson we should draw. We should already know it. | {{Drop|T|hat much is}} obvious: it is not the lesson we should draw. We should already know it. | ||
The lesson is this: if we mistake the map for the territory — if we organise our interests and judge outcomes only by reference to the map we have made, ''we thereby change the territory''. There is a feedback loop at work here. Gradually, by degrees, behaviour will change and territory will converge on the map. | |||
This appears to be excellent news, in the short term, for machines and those who deploy them — cartographers — as it makes their life easier in the “ordinary run” of things. The mountain comes to Mohammed. For those out in the territory, it leads to two kinds of bad outcomes. | |||
{{L1}} Making life easier for devices that work by “algorithm” and see the world in terms of numbers — call these “financialisation machines” — | {{L1}} Making life easier for devices that work by “algorithm” and see the world in terms of numbers — call these “financialisation machines” — thus removing variability does not make our lives richer. It makes them generic. It diminishes the benefit of network nodes that ''can'' handle ineffability — this is good, right, because those nodes — call them [[subject matter expert]]s, or even humans — are expensive, slow and unpredictable. | ||
Now of course we can assign humans to algorithmic roles — where there is peripheral intractability in a | Now of course we can assign humans to algorithmic roles — where there is peripheral intractability in a system, we have no choice — and as the territory redraws itself to the map, we further marginalise that intractability, and can deploy cheaper, more interchangeable humans, and at the limit, we can replace them altogether. | ||
Where the “intractability” is relatively low-risk — for example, interpreting, triaging and responding to unstructured requests from low-value customers, as in a consumer helpline — then techniques like AI can already handle it. It | Where the “intractability” is relatively low-risk — for example, interpreting, triaging and responding to unstructured requests from low-value customers, as in a consumer helpline — then techniques like AI can already handle it. It doesn’t much matter if the AI isn’t much good. (For a complaints line, it is ''ideal'' if the AI isn’t much good — it is a perfect [[accountability sink]]). By agreeing to behave like machines, to be categorised according to numerical terms, to be financialised — we ''surrender'' to machines. If you are worried that your job will be replaced by tech, then this is the way it will happen.<li> | ||
But in those exceptional cases that turn out not to be the ordinary course it leads to bad outcomes in the territory. No better example than the [[Post Office Horizon]] scandal, the internal territory — how managers behave had so closely converged on the map as to utterly lose sight of the territory. As long as the territory was unaligned, scattered, unconnected and could not fight back, this did not matter at all. But the territory has a habit of overwhelming mapmakers who lose sight of their original purpose, which was to functionally reflect the territory. Our [[roll of honour]] refers. | |||
</ol> | </ol> | ||