Contribution
It is the fullest short statement of Soros’s two principles, written for a special issue of the Journal of Economic Methodology that answered it. He claims no discovery (“Of course I did not discover reflexivity”) and names Knight, Keynes’s beauty contest, Merton’s self-fulfilling prophecy and Popper’s Oedipus effect as forerunners. What it adds is the framing of the loop as two functions of one mind interfering with each other, a demarcation of social from natural science built on it, which the postscript retracts, and its application to boom–bust processes and the euro crisis.
Key insight
A thinking participant’s view of a situation is both an attempt to know it and a lever on it. When both run at once, “The independent variable of one function is the dependent variable of the other, thus neither function has a genuinely independent variable”. The facts then stop being an independent test of the view: “if there is interference from the manipulative function, the facts no longer serve as an independent criterion because the statement may be the product of the manipulative function.” His example: “It is raining” is checked against the weather; “I love you” changes what it reports.
Assumptions
- Fallibility. “the participants’ views of the world never perfectly correspond to the actual state of affairs”; theories and overall views are “bound to be either biased or inconsistent or both”. Facts of detail are excepted.
- Reflexivity. These views “can influence the situation to which they relate through the actions of the participants”. It “applies exclusively to situations that have thinking participants”.
- Influence, not determination. “The participants’ views influence but do not determine the course of events, and the course of events influences but does not determine the participants’ views.”
- Two kinds, and a third. Reflexive relations connect subjective aspects (marriage, politics); reflexive events connect subjective and objective (the fiscal cliff, the euro crisis); “self-reflexivity” is a solitary person reflecting on their own identity.
Key results
The essay argues and illustrates; it proves and tests nothing.
- Human uncertainty principle. Fallibility and reflexivity “lumped together”. He then says “it is fallibility that is the key source of Knightian uncertainty in human affairs”; reflexivity “is only one source of uncertainty, albeit a powerful one”. Not every fallibility is Knightian: road accidents and the biases behavioural economists catalogue are statistical.
- Against Popper’s unity of method. Reflexivity breaks the symmetry of prediction and explanation and endangers testing, since initial and final conditions either include participants’ thinking, which is not observable, or exclude it, and then do not fix the outcome. He proposes, “not as an empirical truth but as a methodological convention”, that social science not be held to physics’s standards.
- Postscript. After Beinhocker’s paper in the same issue and a CEU workshop on 8 October 2013, he says the convention “could do more harm than good in the long term” and that “we should study the spectrum rather than attributing reflexivity exclusively to the domain of the social sciences”.
- Feedback signs. Negative feedback brings views and situation together and is self-correcting; equilibrium is “an extreme case of negative feedback, a limiting case”. Positive feedback drives them apart, is self-reinforcing, and cannot go on forever.
- Boom–bust. A prevailing trend and a misconception about it reinforce each other, are tested, survive or fail, reach a “twilight”, and reverse; booms are slow and busts abrupt. Figure 4’s caption stages it A–I. Worked cases: real estate (the misconception “that the value of the collateral is independent of the availability of credit”), the 1982 sovereign-debt crisis (debt ratios “considered objective criteria, but in fact they were reflexive”), the 2007–2008 “super-bubble”, the euro crisis.
- Fat tails. Offered as a partial explanation: dichotomies simplify a reality too complex to grasp, and a market’s switch from one side of a dichotomy to the other is violent.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Where participants think, their views both track and change their situation, so facts are not an independent test of their views | moderate | conceptual argument, with examples |
| C2 | Fallibility, not reflexivity, is the main source of Knightian uncertainty in human affairs | weak | asserted, in one paragraph |
| C3 | Social science cannot meet Popper’s scheme, because the human uncertainty principle breaks the symmetry of prediction and explanation and the replicability of conditions | moderate | argument; softened by the postscript |
| C4 | Bubbles are a trend and a misconception reinforcing each other, with a slow boom and an abrupt bust | weak | the author’s model and narrated cases, no data |
| C5 | Reflexivity explains and predicts financial markets better than the efficient market hypothesis | weak | the author’s own record, “I can testify from personal experience” |
| C6 | The euro crisis was produced by fallibility on both the markets’ and the authorities’ side | weak | historical narrative |
Concepts
- Cognitive function: the participant as observer; causation runs “from the world to the mind”.
- Manipulative function: the participant as actor, “to make an impact on the world and to advance the participants’ interests”; “manipulative” is chosen “to emphasize intentionality”.
- Reflexivity: the interference of the two, when both operate; a “two-way feedback loop connecting thinking and reality”.
- Human uncertainty principle: fallibility and reflexivity together; “much more specific and stringent than the subjective skepticism that pervades Cartesian philosophy”.
- Near- and far-from-equilibrium: the regimes of balanced feedback and of preponderant positive feedback.
Connections
The loop is Merton’s self-fulfilling prophecy (LIT-1398), which Soros cites through Social Theory and Social Structure (1949), generalized to run in both directions and to be mostly unintended. Land’s hyperstition (LIT-1396, NOTE-1378) is the same loop made a programme, where Soros makes it a limit. MacKenzie’s performativity (LIT-1395) is the empirical study of one channel he names in passing: a theory of markets (“if investors believe that markets are efficient then that belief will change the way they invest”) changes the markets. Espeland and Sauder (LIT-1330) is the same loop for rankings. His remark that behavioural economists study only “cognitive fallibility leading to the mispricing of assets” and not the mispricing’s effect on fundamentals is the point CLAIM-426 turns on.
Bearing on the record
- CLAIM-426. The LIT gives the reason the claim keeps visibility as evidence about a future and not as a warrant; the essay bears that out. Soros’s views “influence but do not determine” outcomes, his loops run both ways, and his example of a theory acting on markets is that it changes them “though not necessarily making them more efficient”. Read with CLAIM-426: circulation is evidence about what a judgement will bring about, but not about which way. That supports the claim’s narrowing of CLAIM-425 and says its title should allow the loop to run against the judgement as well as for it.
- LIT-1397’s summary put the uncertainty on the interference of the two functions, “which he calls the human uncertainty principle”. The essay’s principle is fallibility and reflexivity together, and Soros says fallibility is the key source. The summary is corrected.
- TERM-075. Soros names “self-reflexivity”, a person changing their conduct after “asking herself who she is and what she stands for”, as the exceptional case within one subjective aspect. It is the appropriated sense of the attributed self with no other person in it, and he does not develop it.
No instruction for machine-learning practice; nothing for the anthology.
Limitations
- Argued from the author’s experience, not tested. He says so: “my approach is still very rudimentary”, and “Knightian uncertainty cannot be quantified”.
- The demarcation of social from natural science, which Section 3 is built on, is withdrawn in the postscript, and the body was not revised.
- The argument that reflexive statements escape the correspondence test is made by example (“I love you”), and the link to the Liar paradox and Russell is suggestive only.
- Errors and slips.
- “the chess contest between Big Blue and Gary Kasparov”: the machine was Deep Blue.
- “If agents act on the basis of their perfect understanding, equilibrium is far from a universally and timelessly prevailing condition” reads backwards; the argument needs “imperfect”. It may be the web copy’s or the article’s; the journal’s text was not seen.
- Hayek and Popper “recognized in their famous exchange in the pages of Economica (Popper, 1944)”: the reference is Popper’s “The poverty of historicism, II”, not an exchange with Hayek; Hayek’s “Scientism” essays ran in Economica in the same years, so the loose wording may be meant.
- “a steam of objective facts” for “stream”; “set a part” for “apart”.
Open questions
- When is a reflexive loop positive and when negative? Soros says “myriad feedback loops” of both signs run at once and gives no way of telling which will win before it does.