Keystroke Lotteries: A Speculative Essay, Part I

How much work would you do for a lottery ticket?

By Bruce Swanson

bruce.swanson.california@gmail.com

You are at work, at play, at home, in transit. You have a second, a minute, an hour to kill.

You log onto a website where you find a list of typesetting, proofreading, copy editing, or translation projects. Each requires a specific application, such as Word or Excel. All of them are commonly available.

You choose a PDF file containing images of 100 pages of handwritten text of no interest to anyone but the owner, who is offering $100 to have it all typed by midnight.

It's now 9 p.m. You start work.

But within a minute you are too bored to continue. Without hesitation you click the browser window closed and three hours later the full $100 is deposited to your Paypal account.

It's not a mistake.

Some days later, reading about a recent local murder, you realize that you knew the victim and who her murderer might be. But you don't want to risk going public with the information. And you don't trust the police and the courts to safeguard your identity.

You log on to a secure website that allows you and the police to exchange messages securely without them ever knowing who you are. Nor can they ever know unless you tell them first. You give them your hunch.

Within a week your suspect has been arrested. A year later he is tried and convicted.

Upon that conviction you begin privately receiving online a steady stream of big-payout lottery tickets. Because you were the only one to inform the police, you'll keep getting the tickets until one of them wins.

"Everybody, almost, can and will be willing to hazard a trifling sum for the chance of a considerable gain." — Alexander Hamilton

A keystroke, mouse-click, or touch-screen response is a trifling sum of work that everybody on the Web, almost, would be willing to hazard for the chance of a considerable gain.

In the hypothetical situation described above, you began to type out handwritten text that had been scanned into a PDF. But so did hundreds, perhaps thousands, of others. Working simultaneously, many quitting just like you did as others joined in, the 100-page project was completely typed with time to spare. The funder of the lottery not only got a great deal at a page-rate of one dollar, the work was finished much faster than could have been done using conventional pay-arrangements. And the capacity to automatically count and record the total number of different keystrokes you and everyone else typed all but guaranteed a numerical keystroke-consensus per written character.

In effect, the document was proofread as it was typed.

That has never before been economically feasible, and the implications of that are what makes the keystroke-lottery concept revolutionary.

You can see why someone would fund such a lottery: programmatic redundancy is always best in data entry and proofreading. It is often good in copy editing, and might even be good in certain kinds of translations, such as non-stylistic, non-literary efforts where speed and legal intelligibility alone are paramount. Redundancy is the principle behind crowdsourcing, defined by Wikipedia as the act of outsourcing tasks, traditionally performed by an employee or contractor, to an undefined, large group of people or community (a "crowd"), through an open call.

Because the total amount of money offered ($100) was rather small as compared to conventional lotteries, during the auction (which was for a place in the site's work-queue), the owner of the pages guaranteed a winner and specified no lower limit on the total quantity of keystrokes per person needed to win. That means to play, you needed only to type the minimum number of keystrokes sufficient to establish a context on the page, verifiable by enough others typing the same keystrokes to represent the same handwritten letter or digit. In this example, had the document started with A, then that single keystroke would have earned a chance to win the payout.

The thesis of this essay is that growing ease and universality of visual communication must and will eventually turn chance itself into compensation for certain distinct kinds of online work. That chance — to win money for minimal work—will guarantee access to a world-wide virtual crowd of Web users whose sheer numbers will free them to participate as little or as much as they want, when they want, hazarding their time and typing in the lottery associated with the project on which they have chosen to work. Group keystrokes from around the world would repeatedly wash in real time like a digital aurora over a given document, page, paragraph, sentence, or Captcha-like word-picture in play at the moment. Conventional work-schedules and commitments and piecework-schemes would be unnecessary.

A subtle but important point to be understood is that each typist would be working not specifically to be accurate, but rather to achieve consensus with others. Because only a consensus-validated keystroke could win, for the typists there would be no correct keystrokes per se, only consensus-validated ones. The document would be intently followed to achieve that consensus as quickly and easily as possible, and only documents that showed a reasonable chance for the money offered of achieving a consensus would be accepted by the crowd. Thus, keystroke lotteries could just as well be called consensus lotteries.

The need for consensus also dictates that typists will select only projects in which they have the necessary expertise. They would not be maximizing their chance of winning if they were not so selective. Thus, projects requiring a critical amount of recondite knowledge would cost more to fund because fewer typists would be interested in gambling their time on them. As with jobs like the one pictured above, documents trafficking in obscurity would have to radically increase their offered winnings to generate enough interest to provide meaningful consensus.

That point understood, gambling would quickly go to work. As in the hypothetical case above, to make more interesting a small payout for snippets of typesetting or proofreading (crap in printing-parlance), an appeal would have to be made to immediate gratification with an instant guaranteed-winner upon project completion. As is the case in traditional lotteries, larger payouts wouldn't require a guaranteed winner per drawing. (Insurance, funded by micro-lending members of the public who prefer that form of gambling, might be offered to those needing higher payouts than they are personally willing to risk.) Crowdsourcing would mean crowdpaying, and the pay would be the lottery tickets earned when individual eyes, brains, and fingers acted in sync with a crowd changing by the second, yet remaining unchanged in its specific purpose.

As with any group working together, there would have to be a mechanism for a single typist or minority of typists to forge a new consensus in the face of the majority. That mechanism would probably be a kind of side-bet. A typist, realizing that the document contained a factual or stylistic error, might offer a correction and then flag it. That flag would attract the attention of others looking for consensus. So the dilemma for those other typists would be whether to validate the flagged change with their own keystrokes, or ignore it. They might ignore it at their peril, if newer typists logging on saw the flag and decided to back it up. Participants might even bet their earned tickets on the outcome of a flag. Typists might allow a public record of their corrections, thus lending credence to their flags. Thus would typeset-led proofreading morph into copy-editing, a mechanism that might also make group translations possible (of which, see below).

Because all this activity would be taking place in a digital environment, it might be possible to select projects that offer odds modeled on popular conventional lotteries. This would allow participants to learn the practical meaning of the odds that such lotteries offer, but without paying cash for the lesson. It might also be possible to set parameters to price participants' typing as an hourly wage up to the amount that they would have paid for a conventional lottery ticket. The program would notify them when they reached their pre-set limit, at which point, to the keystroke, they could quit. In any case, as many chances beyond their customary daily or weekly ticket-purchases as they would earn at their keyboards, most people would still never win a significant payout — even as they watched a news-ticker roll across their monitors announcing the names of winners around the world who had.

So a keystroke lottery could be educational enough — California Lotto revenues are supposed to help education — to make conventional cash-purchases of lottery tickets unappealing to larger and larger numbers of players. More and more of them might switch to keystroke lotteries, or even — the more perceptive among them — quit lottery-gambling entirely. That said, few players have ever won lotteries, but that has always been enough to regenerate interest for a given payout. A chance for the money, as always, is the attraction. Quitters would return to the fold.

The Wikipedia Example

If Wikipedia's business-plan had first been posted online for general review, most readers probably would have accepted its theoretical potential, but not its chances for practical success. What most surely would not have predicted is the spontaneous growth and increasing complexity of Wikipedia's participatory culture — that growing body of knowledge, experience, foresight, governance, and enforcement that have come to characterize the experience of using and maintaining it. (Wikipedia may be the first and only effective form of mass communism the world has seen thus far.)

Because gambling is (arguably) based on psychology as much as mathematics, world-wide keystroke lotteries probably couldn't be computer-modeled to the point of foregoing all funding for further development in the manner of, say, cold fusion should any given model indicate failure. Only by trying it in real time with real people winning enough real money could its technical feasibility and potential popularity be determined. Until that happens, one sees nothing in the idea of keystroke-lotteries that would cause any reasonable person to state outright that they couldn't work, that a culture similar to Wikipedia's couldn't arise.

But there is another fundamental point: Wikipedia offers no compensation because it doesn't have to, being viscerally attractive to huge numbers of volunteers. By contrast, the website Distributed Proofreaders (for example) also offers no compensation, but isn't remotely as popular as Wikipedia because the work it offers isn't creative at any level and is thus fundamentally uninteresting to too many people. The prospect of proofreading a given book by keyboard may interest you personally, but without the prospect of financial reward it won't interest the general public. This is why DP is a useful but tiny part of the Internet, its momentary slashdotting in 2002 notwithstanding. But putting a gaming front-end on equally uncreative data-entry work could put keystroke lotteries on par with Wikipedia in popularity and impact, attracting millions of disinterested strangers willing to do their best working together solely for the chance of winning a payoff. If the result efficiently lowers cost and reduces production time — in other words, if it works — who could object?

The federal government might object. It follows the laws set by Congress, so that body's gaming-industry contributors could be expected initially to fight any lottery system not already fully described and permitted by law. More locally, there are states with lotteries and states without them. It's possible that both would oppose a keystroke lottery, either from fear of a loss of state-lotto earnings or out of opposition to any legalized form of online gambling. In that event, the first keystroke-lotteries inevitably would move offshore. On the other hand, all interested parties might co-opt keystroke lotteries as a promising revenue-source.

Legal objections notwithstanding, keystroke-lotteries are surely well-within our technical capability. Economics are another matter: we must consider the marginal cost of each additional lottery-typist versus one staff proofreader or freelancer hired by the hour who could be counted on over time to miss more mistakes than a crowd would. That marginal cost would no doubt vary from project to project, but in general it probably would rise to a prohibitive point, reflecting the fact that once you've gotten the crowd to a certain size for a document of a given perceived complexity, the wisdom of that crowd would not be economically increased by adding any more people to it. But when would that additional typist would become prohibitive, and at what point would a game not need a guaranteed winner? An auction would provide the answers.

As for collusion, it would surely be uneconomical at any scale. One would have to organize a ring of typists and split the (rare) winnings with them, all the while knowing that the keystrokes of enough unorganized players would swamp the conspiracy. Also, typists might (in some lotteries anyway) be able to see what other typists are doing in real time. Software could also detect suspicious patterns and block IP addresses. Endless waves of cheaters would inexorably realize that they might just as well spend their time doing what everyone else is doing, given the tiny chance of winning. As with any lottery, tiny input, tiny chance; much larger input, very slightly larger chance. Cheating would add to the real work necessary to theoretically win, and would probably have to limit itself to the smaller-payout lotteries with fewer participants. But even then, keystroke lotteries would still be premised on the little amount of work required of any one person to participate. So protective levels of participation might be a given even in lotteries with small payouts. As a further measure, projects might be graded by a statistical measure showing the theoretical effort required to cheat it, a measure that would certainly be reflected in the project's bidding level.

"You can get it cheap, fast, or right—but only two out of three."

— Print-shop maxim

Small, neighborhood convenience-stores can charge premium prices while remaining competitive with bigger-box stores located nearby. Likewise, keystroke lotteries would not necessarily be less expensive than work done conventionally. They might even be more expensive, especially before they became ubiquitous enough to induce sufficient competitive bidding among massive numbers of participants. Instead, keystroke lotteries would be a radically cheaper and faster way to deliver the highest practically attainable levels of quality for a given volume of work within a given deadline. In other words, attaining such quality on time under traditional methods would be radically more expensive than using a keystroke lottery. Understanding this distinction is critical. Initially not everyone would be willing to pay the premium for that quality as a function of time. But those who were willing would get their job done cheaper, faster, and righter than ever before. Cheaper, faster, and righter would eventually define cheap, fast, and right. The above-quoted maxim would be overturned as a consistent principle for the first time in the history of printing.

It obviously remains to be seen how quickly the public would understand exactly what a keystroke lottery isn't. The key selling-point would have to be consensus, a requirement limiting creativity although not (as indicated above) prohibiting it entirely. Thus keystroke lotteries may not be expandable to polls and advice-giving, formats not based directly on a template. However, it might apply to certain forms of creative copy-writing, such as found on Trada.com. There, crowds of ad-writers try to write ads that merchants want written but don't have the time or expertise to write themselves. That merchant might specify a maximum price-per-click of $1.00. A writer in the Trada crowd who can write the merchant an ad that gets clicks at below a dollar is paid the difference (minus the Trada fee). It may be that that kind of competitive writing bounded by strict space and content limitations (keywords and creative URLs) will attract people willing to write complete but short Google ads, with the money they collectively earn going into a pot that one of them would win.

Another recent try at crowdsourcing is Gigwalk.com, in which smartphone users are paid to provide local bits and pieces of mapping or pricing information too expensive to obtain conventionally. The rate starts at $3 per gig. But how much work would someone do for that three dollars? And how much work would they do if a lottery ticket for a thousand-dollar pot cost $3, funded by pooling a day's or week's or month's worth of $3 gigs? To ask that question is to repeat what I initially asked at the top of this page: how much work would you do for a lottery ticket?

For perspective on the question of overall acceptance, imagine that keystroke lotteries were first in the history of lottery gambling. Then someone came along and proposed the following: instead of having people access to work-games from the convenience of their homes, offices, and cellphones, why not have them trek to the corner liquor store? And instead of having people gambling on the outcome of real work, why not have them gamble on colored images of fruit? And pay with their own cash for the privilege?

Would the general public reject keystroke-lotteries even if there was ready money available to be won merely by logging on and typing for a few seconds or minutes at a computer where they already happen to be, and are going to be, hour after hour, day after day? It doesn't seem likely, when a single keystroke or mouse-click could (statistically speaking) win, and the wider public's general forbearance would — to its quick realization — merely advantage a comparative handful of early-adapters.

Translations

"I have at last discovered the right way to translate Onegin. This is the fifth or sixth complete version that I have made. I am now breaking it up, banishing everything that honesty might deem verbal velvet and, in fact, welcoming the awkward turn, the fish bone of the meager truth." —Vladimir Nabokov, The Nabokov-Wilson Letters, 1940 –1971.

Nabokov believed that no honest translation could aspire to, or pretend to achieve, artistic unity with the language of the original work. Although his own novels were conventionally translated for commercial reasons, he put his philosophy rigorously into effect in his own literal translation of Eugene Onegin, by Alexander Pushkin. Nabokov's model has not caught on with the public, and today there is considerable discussion about what the future holds for professional translation-services. Even for the strictly monolingual (such as myself), it seems reasonable to suppose that a document translated by a keystroke lottery would create enough consensus on which to base a lottery while simultaneously producing significant differences (fifth or sixth complete version). But given the digital environment of a word processor, those differences could be retained as separate documents, or expressed by color coding, typefaces, or type sizes (take your pick); or abstracted statistically and displayed in charts and graphs. Degrees of consensus could be indicated by a number on a scale, and documents could have that number appended to them. All this would be an advantage where the highest degree of "accuracy" (I think Nabokov would approve of the quote marks) would be needed as quickly and/or cheaply as possible. It should be remembered that the whole point of the process is consensus and nothing else, so individual expressions of style might persist as typists realized a non-literal construction that was also likely to suggest itself to other typists, to the point of flag-betting for it or against it. Nabokov followed his own sense in deciding when to finish his book, a sense no doubt informed at least in part by the dictates of time and money. Those same dictates would govern typist-translators, albeit within a radically different format. Keystroke-lottery translations would not likely be popular as literature, but literature would not always be wanted or needed, either by the writer or the reader.

Computerized translations will continue to improve in all-around "accuracy" (quotes again), yet for documents of legal importance especially, human input will still be needed for a long time to come. I think that more and more translations via lottery will become contrary betting-games—that typists will scour documents for things to bet against, as described above.

The Question Answered

How much work would you do for a lottery ticket? Very little — at any one time. That's the whole idea.

Next: Part II -- Informant Lotteries