An Ethical Debate about Robots, Shorter Workweeks and Unemployment

I was delighted to receive a cordial and thoughtful reply from Omar al-Ubaydli to my open-letter blog post to him about his "counterpoint" piece on shorter working time and unemployment. The purpose of this post is both to reply to Omar and to recap and contextualize the exchange to make it accessible to the general reader.

A few days ago, I was writing a blog post about the significance of the "first use" of the expression, "lump o' labour" in an 1843 novel by John Mills. Mills's use of the expression would appear to faithfully reflect what its common meaning and context of usage would have been to ordinary working people in the mid-19th century -- at least there is no reason to suspect otherwise. That meaning would have been concrete and definite: the expenditure of just so much effort to perform a given task. The context would have been the idea of a appropriate, proportionate compensation for the amount of effort expended.

In the midst of writing that post, I heard about the point-counterpoint exchange on shorter working time and employment between Dean Baker and Omar al-Ubaydli. Dean's piece was titled Shorter Workweeks Will Defeat the Robots and Omar's was titled, Shorter Workweeks Will Increase Unemployment. Omar's counterpoint waved the red flag of the fixed amount of work:

A glaring red flag is how simple the proposed solution seems to be: Proponents of work-sharing believe an economy requires a fixed amount of work to be performed by a limited number of people. High unemployment, they contend, is due to allocating too many hours to current employees. A more equitable redistribution of work hours, according to this logic, may diminish unemployment: Instead of Alex and Chris working 45 hours a week and Jo being out of a job, each can work 30 hours, eliminating unemployment.

... 

In contrast with an elementary work-sharing analyses, the real world reveals that there isn’t a fixed amount of work to be done. The total demand for labor depends upon how it’s restricted or divided.

Much of the discourse about robots -- and, before that, about automation or machinery -- has revolved around the question of the amount of work that will be left for people to do. This is asking the wrong question. But it is the question they want you to ask.

"If they can get you asking the wrong questions," Thomas Pynchon warned in Gravity's Rainbow, "they don’t have to worry about answers." In this case, the question they don't want to worry about answering has to do with proportions, not quantities.



A quantity is not the same as a proportion. A proportion of anything is always in comparison to something else. The quantities in a relationship can change while the proportions remain the same -- think of doubling the recipe for a chocolate cake. But the proportions change if some quantities change while others don't -- or while others don't change proportionally.

This is where the ceteris paribus assumption, "all other things remaining the same," runs into a paradox. If "x" changes and all other things remain the same, then the proportions between "x" and everything else change and thus ceteris is definitely not paribus. The illustration on the right, below, shows what happens to Leonardo Da Vinci's Vitruvian man when his head is enlarged 25% "ceteris paribus":

Proportionality was central to both Maurice Dobb's and A. C. Pigou's objections to the fixed Work-fund (lump-of-labor) fallacy claim and to Charles Beardsley 1895 refutation of John Rae's homologous claim:

"What [economists] intended to say was that a restriction on the supply of labour could not increase aggregate earnings... Such restriction can, however, increase wages in proportion to the worker's expenditure of energy and his 'wear and tear,' and it can increase Relative Wages, or wages as a proportion of the total social income." -- Dobb 

"...the reduction of the normal hours of work can create employment for them on one condition and on one condition only; namely, that these devices succeed in rendering the labour and capital of the rest of the community more effective in production. ... The reduction of the normal hours of work will do it or will fail to do it, according as the extra leisure increases the workers' efficiency more or less than in proportion to the reduction of hours." -- Pigou

"...it is not certain that the demand for labor increases in proportion to the increase in the supply of capital and machinery. By the adoption of shorter hours for work, the available supply of labor in relation to the demand for it would be reduced, and the price of labor tend to  rise. But, if the higher price checked the growth of capital,  it would hardly check the demand for labor to an equal degree." -- Beardsley

Note that these are not the same argument. But they all rely on proportions, not simply quantities. Combining Dobb and Pigou, one could argue for a shift in relative income shares, an increase in leisure and productivity and an increase in employment without the magnitude of that increase in employment being the sole criteria of policy benefit.

John Graunt started it all of in 1662 with his admonition about a certain proportion of work "…if there be but a certain proportion of work to be done... then to employ the Beggars about it, will but transfer the want from one hand to another…" Graunt’s concern with proportion was a legacy the influence of Aristotle’s Politics and his Ethics on early modern thought. Thus Graunt's political arithmetick was explicitly concerned with understanding the political. In contrast, much of the political economy of the 19th century was concerned with subordinating the political to the expansion of trade.

I sent a tweet to Omar to let him know about my commentary on the early usage of lump o' labour in which I had characterized his red flag remarks as yet another rendition of the stale colonial discourse that dismisses concerns about technological unemployment as the ignorant ravings of a "bunch of goofballs." Omar asked me to elaborate on my point, which I did in the blog post, mentioned previously. Midway through, I pivoted from disputing his claim about what proponents of work-sharing believed to proposing an ethical debate or dialogue according to the principles outlined by Anatol Rapoport in Fights, Games and Debates:

1. The ability and willingness of each participant to state the position of the opponent to the opponent's satisfaction (exchange of roles)
2. The ability and willingness of each participant to state the conditions under which the opponent's position is valid or has merit (recognition that any position whatever has some region of validity)
3. The ability and willingness of each opponent to assume that in many respects the  the opponent is like himself; that is to say, that a common ground exists where the opponents share common values, and each is aware of this common ground and, perhaps, of the circumstances which have led the opponent to the position he holds (empathy)

I initiated the process by summarizing his position and invited him to do the same with mine. For details, see my letter to Omar and his reply. One additional strand of context before I respond to Omar's reply. It was in researching the "zero-sum fallacy" that I encountered Rapoport's discussion of ethical debate.

Rapoport did laboratory research on game theory and was a critic of its misuse. He wrote an article for Scientific American on "The Use and Misuse of Game Theory." His principles for an ethical debate were based explicitly on his deep understanding of the dilemmas inherent in non-constant sum, multi-player games and the perils of treating real world dilemmas as if they were games. Rather than go into a rambling digression here about game theory and ethical debate, I'll just cite the links for the "zero-sum foolery" series:

1. Game theory gamesmanship,
2. Doomsday climate machine,
3. Forecast Factory,
4. Wage prisoner's dilemma,
5. Is the road to hell paved with pareto improvements?

Number five is kind of an epilogue. There was one more piece in the series that never got posted because it remains unfinished. It was titled "It's Mercantilism all the way down." What follows is cannibalized from that draft. It wasn't really about mercantilism, per se. It was about polemic, dialogue and theory and the differences between them. The Sandwicman has been fighting polemic with polemic for nearly twenty years to no avail. I wanted to try to sketch out a dialogic alternative.

In his 1948 article, "Dialectical Materialism and General Semantics," Rapoport outlined the technique used by Lenin in his polemic against "Machism" in Materialism and Empirio-Criticism. Lenin sought to discredit his philosophical adversaries by establishing that there are only two consistent world views and validating his own perspective, materialism, as having been "derived from self-evident facts with inexorable logic."

Although the adversary's view, idealism, may also be logically derived, it proceeded from dubious assumptions that are incompatible with the allegedly self-evident facts. From this polemical perspective, science "consists in carrying out a relentless struggle" against the proponents of the erroneous world view.

Having established that there is an "unscientific" and a "scientific" point of view -- a myth and a reality -- the task for the conscientious polemicist is then to demonstrate that the opponent does indeed make the erroneous assumptions from which the unscientific, mythical view is logically derived. Lenin took elaborate pains to show that his adversaries believed what he said they believed in spite of any protestations to the contrary.

Rapoport contrasted Lenin's meticulousness with "modern efficient smear methods, such as defining subversive activity to be attendance at a concert by Paul Robeson." The problem with polemics, in Rapoport's view, is that there are no winners "in the sense that no evidence exists of significant conversions to this or that point of view ever having taken place as a result of polemic." This raises the possibility that the function of polemic may often be to forestall, rather than to "win" a debate.

Classical political economy -- or at least some key tenets of it -- may be usefully viewed as a polemic against mercantilism. The polemical strategy is to establish the correctness of the polemicist's view on the grounds of the error of the opponent's (supposedly) contrary view. The opposite of false is true. If that guy is wrong, I must be right! Whether the two positions are, in fact, "opposites" is a question that is best either evaded or obscured...

In Fights, Games and Debates (1960), Rapoport proposed a set of procedures for carrying out an ethical debate. The first rule for such a debate is that each party must present the opponent's case to the opponent's satisfaction before presenting their own case.

The second step is to identify specific situations in which the opponent's position is correct. This is possible even in extreme situations. Rapoport gives the example of the claim that "black is white." One can agree that in a photographic negative, black is indeed white.

Third is to tease out common ground and potential agreement on values.

Finally, in presenting one's own case, one may proceed to point out the circumstances in which the opponent's claim is not applicable. The purpose of these procedures is to encourage each participant in the debate to identify with the opponent by empathizing with the thought process that produced their contentious position.

Of course there is no guarantee that the opponent will follow the rules of an ethical debate. Rapoport later recalled an incident with an associate director of the Institute of Advanced Studies in Vienna. He began by presenting what he viewed as the associate director's perspective,  "you think that I did something just in order to embarrass you." The associate agreed that this was indeed what he thought. When asked to state Rapoport's perspective, though, he replied, "Well, you have just stated it. You said yourself that you did it to embarrass me, didn't you?"

Rapoport might have retorted, "You think that I admitted that I did something just to embarrass you." To which the associate could have responded, "Yes and you just confirmed that you did admit it!" And so on, in Monty Python argument clinic fashion.

Thankfully, Omar's response was more responsive than that of the associate director. He admitted it was sloppy "to imply that all the proponents hold a certain position when the reality is clearly that only some of them do." He went on to outline the constraints he was under in writing the piece and suggested that it would be better if there was open disclosure of these guidelines to avoid confusion. 

I regret that Omar didn't fully take up my invitation to follow the Rapoport principles and summarize the position of his opponent, pointing out where it is valid. I had suggested Pigou and Dobb as succinct versions of the opponent's position. In a nutshell, Omar's summary of his opponent's position consisted only of the admission that not all proponents of work-sharing believe there is a fixed amount of work to be done. What it is they do believe remains opaque.

I would again invite Omar to state the position of the opponent to the opponent's satisfaction and to state the conditions under which the opponent's position is valid or has merit. In the meanwhile, I will respond to two points that Omar raised regarding my open letter: that my summary of his position "omits the most important part, which is what the data from Europe show", that I regarded his contribution "as verging on the conspiratorial, or reeking of ignorance" and that I should refrain from accusing contributors of lying.

With regard to the last comment, I didn't accuse anyone of lying, so the advice is nugatory (unless someone has access to frenzied tirades I compose in the wee hours and have better sense than to dispatch). I did, however, call the assertion, "proponents of work-sharing believe an economy requires a fixed amount of work" untrue. There is an important difference between lying and uttering an untruth. If one says something that is not true in the mistaken belief that it is true, one is not lying.

Reeking of ignorance? I quoted John Stuart Mill, "He who knows only his own side of the case, knows little of that." One can be intelligent and well-educated and still know "only his own side of the case." This may be more a matter of hubris than ignorance. 

And finally, "omitting the most important part." My summary of Omar's position didn't include his reference to the data from Europe because what I was taking issue with was his contention that proponents of work-sharing believe there is a fixed amount of work. The data from Europe do not in any way reflect on either Omar's contention or what the proponents' believe. Data may even "be consistent with" a false hypothesis. Such agreement would not make the hypothesis true.

Getting down to the brass tacks of the data, the research findings regarding working time are indeed inconclusive, as are the conclusions having to do with deregulation of labor markets, which Omar favors and for which he wrongly claims the international data provides conclusive evidence. 

Richard Freeman argued that there were two reasons that the debate over labor market liberalization was inconclusive. First, many researchers hold strong priors about the desirability of unregulated labor markets and "let their priors dictate their modelling choices and interpretation of empirical results." Second, the cross-country aggregate data is "too weak to decisively reject strong prior views or to convince those with weaker priors." ("Labour market institutions without blinders: The debate over flexibility and labour market performance," International Economic Journal, Vol. 19, No. 2, 129–145, 2005). Dean Baker has also carried out research in this area. 

I reviewed the research on work time reduction and employment in both 2000 and 2007 and came to a similar conclusion as Freeman. There are too many confounding variables to make cross-country comparisons about the relationship between work-time reduction policies and unemployment without heroic assumptions about how to control for the confounded things. I was, incidentally, appalled by the models economists used to estimate the effects of work-time reductions. Several models I reviewed assumed that the given hours of work were optimal for output prior to the implementation of work-time reduction policy and thus by assumption, policy would have a negative effect on hourly productivity and output. I could go into a technical discussion of why this assumption is not only inept but reeks of ignorance...

But I digress... bad model, good model; weak data, strong data, conclusive, inconclusive -- analysis of the aggregate employment data doesn't offer any insight into what proponents of work-sharing do or do not believe.