[This post provides a followup to the chapter "Addicts in the ED" in my book "Game-Changer: Game Theory and the Art of Transforming Strategic Situations".]
In “Game-Changer” (published Jan 2014), I discuss the problem of rampant opioid misuse and abuse from a game-theory point of view. At the end of the chapter, I predicted the future direction of the strategic ecosystem around opioid prescriptions in the Emergency Department, that hospitals would enter a “race to toughness,” adopting new strategies to make it more difficult for drug-seekers to get opioids in their own facilities, without addressing the underlying problem of opioid dependence.
This is now happening. Earlier this week, NPR reported on the rise of programs to deny narcotic painkillers (opioids) to patients who are found to be abusing the system. See “Emergency Rooms Crack Down On Abusers Of Pain Pills.”
Such programs help hospitals by reducing the burden of serving drug-seekers (freeing up resources to serve other patients) but also improve the bottom line as drug-seekers are not profitable patients from the hospital’s point of view: “Because many of these narcotics-seekers lacked insurance, eliminating their repeat ER visits saved [San Juan Regional Medical Center in Farmington, New Mexico] about half a million dollars a year, enough to pay for about six full-time nurses in the emergency department.”
Unfortunately, such policies do little to address the fundamental problem of opioid misuse and abuse. Much better would be to adopt best-practice policies, such as those spearheaded by Yale’s Gail D’Onofrio, to steer opioid-seeking patients toward programs that can help them control of their pain and their drug dependency. I’m grateful to Dr. D’Onofrio for the help she provided when I was researching how to change the game of opioid misuse back in 2012.
It’s about time that her ideas received the wider embrace that they deserve. See “Why Not Start Addiction Treatment Right In The ER?”.
Yesterday morning, I was surprised to learn (in this Nature.com interview) that Google Scholar lacks an application programming interface (API) because, in the words of Google Scholar co-founder Anurag Acharya, academic publishers allow Google “to scan all [their] articles, but not to distribute this information to others in bulk”. I’m extremely grateful for all the things that Google Scholar allows me to do (see e.g. this paean by Jonathan Eisen), but Scholar could supercharge scholarly research in new and powerful ways if it had an API. (Just consider our experience with Google Maps. What began as a way to find directions is now much more than a map, powering all sorts of unexpected applications.)
In yesterday’s post “Changing the game of scholarly search”, I discussed ways in which a new scholarly-search platform [which I called "Delphi"] could successfully enter the market and build the strategic ecosystem necessary to support programmable search WITHOUT Google’s help. In this followup post, I will discuss how Google Scholar itself could change the game for the better. (I’m motivated to do so by the fact that someone at Google Scholar @GoScholar seems to have followed — and favorited – some of the Twitter back-and-forth on this topic yesterday.)
[Please read yesterday's post before continuing.]
OPT-IN vs. OPT-OUT
A key feature of my proposed strategy for Delphi was that it set itself up as an “opt-in platform,” namely, that it only index and offer programmable search scholarly works from outlets that explicitly grant permission (and perhaps build out some interconnection functionality) for programmable search of its content. The reason why I chose an opt-in model for Delphi is that opting in gives the very FIRST publishers who join the platform a differentiating advantage. Drawing in the very first adopters is always a challenge when launching a new platform, and I figured that programmable search could be an attractive differentiating feature … enough so that some academic publishers might be willing to make their works Delphi-searchable in order to set themselves apart from the crowd.
But if Google were the one to launch Delphi — as “Google Scholar API”, much as it has “Google Books API” — there is no strategic reason to make the system “opt-in”. Indeed, the natural move for Google in this space would be to launch an opt-out API platform, giving academic publishers the choice not to have their content programmably searchable, but allowing all scholarly content in the public domain to be automatically accessible for search through the API.
RACE TO THE TOP
Once Google Scholar API is up and running, the same “race to the top” dynamics will naturally unfold as I laid out in Delphi’s case:
**1** scholars will naturally use the Google Scholar API to complement their existing work, even if many of the best academic publishers opt out
**2** papers published in API-accessible outlets will get more exposure and more cites, giving these outlets an advantage in terms of attracting good submissions and rising in the journal rankings
**3** more and more academic publishers will choose to adopt Google Scholar API, making it a more and more powerful scholarly-search platform as time goes on
This really is a case where, if Google builds it, they will come.
Google has done so much for scholarship already. But with an opt-out Scholar API, Google can change the game of scholarly search for the better … again!
In a telling exchange with Nature.com, Google Scholar co-founder Anurag Acharya cited academic publishers as the roadblock behind Google Scholar’s failure to offer a programmable interface.
Q [Nature]: Many people would like to have an API (Application Programming Interface) in Google Scholar, so that they could write programs that automatically make searches or retrieve profile information, and build services on top of the tool. Is that possible?
A [Acharya]: I can’t do that. Our indexing arrangements with publishers preclude it. We are allowed to scan all the articles, but not to distribute this information to others in bulk. It is important to be able to work with publishers so we can continue to build a comprehensive search service that is free to everybody. That is our primary function, and everything else is in addition to this.
In other words, don’t hold your breath waiting for Google to add powerful new functionality to Scholar anytime soon.
The good news is that this glaring weakness, that Scholar lacks a programmable interface, creates an opening for a different sort of scholarly-search engine that can both help fill this void and put pressure on academic publishers to open up their content.
Here’s how that might work:
STEP 1 – Scholar alternative [“Delphi”] is built
An alternative scholarly search engine – let’s call it “Delphi” – is introduced that aims to provide the programmable searchability that Google Scholar lacks. To be clear, Delphi is NOT designed to compete with Scholar. People will continue to go to Scholar whenever they have a specific scholarly search. Delphi’s advantage will instead lie in allowing scholars to conduct programmable searches on the subset of scholarship that Delphi indexes.
STEP 2A – Open-access journals adopt Delphi
Unlike Google Scholar, which attempts to automatically scan all scholarship, Delphi is an “opt-in” platform that only indexes a subset of scholarly outlets. In particular, it only indexes the articles of academic journals/repositories that have taken steps to opt into the Delphi platform. At first, only open-access outlets will choose to opt-in.
STEP 2B – Scholars adopt Delphi for programmable searches
Scholars have an incentive to use every tool at their disposal, no matter how limited, as long as it makes them more productive. Once an API exists to automatically search even a fraction of relevant academic articles, scholars will have an incentive to use that API to learn about recent papers that are of interest to them.
STEP 2C – Delphi creates new distribution channel, encouraging entry of innovative new journals
By enabling automatic scholarly search, Delphi cracks the traditional “chicken-and-egg problem” of academic publishing, that an outlet can only be successful once scholars in a field all know about it. With Delphi, a newly published paper that is highly relevant to some scholar will get their attention, even if they have never heard of the outlet where it has been published, as long as that outlet is indexed.
STEP 3A – Authors prefer to publish in Delphi-indexed publications
Publishing your work in a Delphi-indexed publication will get your research noticed by others working in your field. As a result, Delphi-indexed publications will generate more citations. This extra visibility will boost scientists’ careers and reputation, giving authors a reason to favor outlets whose publications can be automatically searched.
STEP 3B — More and more publishers opt to be indexed by Delphi
As Delphi-indexed publications receive more visibility and more citations, academic journals that submit to Delphi indexing will see their own reputations improve. The resulting “race to the top” will put pressure on academic publishers who insist on boycotting the Delphi platform.
STEP 4 – Paywall publishers change their position; now support a Google Scholar API analogous to Google Books API
The rise of Delphi will put pressure on paywall publishers to offer open programmable search of their content. Rather than joining Delphi’s relatively radical open model, however, these publishers’ natural move will be to retain as much control as they can, by pushing Google Scholar to change and become more like Google Books.
Google Books offers free snippets, fully searchable content, and a programmable interface, while still allowing publishers to charge for full access to content. Allowing a similar amount of access to their content would “take the heat off” paywall publishers by reducing their citation disadvantage relative to Delphi-indexed outlets, while still allowing them to keep their paywall-based business-model intact.
The end result?
In the end, paywall publishers will be compelled to loosen their grip, a bit, on the scholarship that they control. More than that, however, an openly-searchable research platforms like Delphi can empower an alternative publication model that leverages programmable search to distribute the latest scholarship and advance the frontier of knowledge faster than ever before.
I think that’s a vision worth working towards.
Update: This post featured on ICIS blog (Innovating Communication in Scholarship)
“Science for the People” is an outstanding weekly science-related podcast. In July 2014, they ran a show featuring “Game-Changer.” In our fun and wide-ranging conversation, host Rachelle Saunders and I discuss everything from:
- why game theory does NOT assume rationality
- ways to escape the Prisoners’ Dilemma
- the dubious origins of the 1970s cigarette advertising ban
- why diamonds AREN’T forever
- changing the game of neglected diseases
- changing the game of opioid misuse and abuse.
- July 2014: “Beating Amazon at Its Own Game” by David McAdams, Pittsbugh Post-Gazette op-ed, July 4th, 2014.
See the Game-Changer File on “Showrooming” for more on this topic.
- June 2014: “Could a top-two primary tame the Tea Party threat?” by David McAdams, The Monkey Cage (Washington Post blog), June 16, 2014.
In this op-ed for the Washington Post’s political-science blog, I argue that Tea Party success in unseating Republican incumbents could prompt conservative statehouses to adopt a so-called “top-two” primary system.
- March 2014: “Making a Market for Overdose-Halting Drugs” by David McAdams,New Jersey Star Ledger op-ed, March 25th, 2014. (Also appeared in Raleigh News & Observer.)
In this op-ed, I propose taking a page from what works to ensure an adequate supply of electricity (so-called Forward Capacity Markets) to ensure an adequate supply of essential generic drugs, such as the opioid-overdose lifesaver called naloxone.
Radiolab ran a fantastic segment this weekend on the game theory of the “Golden Balls” game show. Listen at http://www.radiolab.org/story/golden-rule/. In the show, contestants must choose whether to “split” or “steal”: if both split, they split the money equally; if one splits and one steals, the “thief” gets everything; and if both steal, neither gets anything. The drama of the show is that, usually, both contestants attempt to convince one another that they plan to share, but then steal anyway. In one famous episode, however, Nick Corrigan took a very different tack when playing with Ibrahim Hussein, telling him:
“Ibrahim, I want you to trust, 100%, I’m going to choose the ‘steal’ ball. I want you to choose ‘split’ and, I promise you, I will share the money after the show … If you do ‘steal’, we’ll both walk away with nothing.”
The Freakonomics blog discussed this same episode back in April 2012, under the title “UK Game Show Golden Balls: A New Solution to the Prisoner’s Dilemma” http://freakonomics.com/2012/04/25/uk-game-show-golden-balls-a-new-solution-to-the-prisoner%E2%80%99s-dilemma/
The most important thing about the Split-or-Steal game, however, is that THIS IS NOT A PRISONERS’ DILEMMA. For this to be a Prisoners’ Dilemma, each player must have a dominant strategy to steal, i.e. stealing must make them strictly better off regardless of what the other player does. In the Split-or-Steal game, however, each player is (financially) indifferent between splitting or stealing when the other person steals. (Whether we both steal or only the other player steals, I get nothing.)
What this means is that whether a player wants to split or steal, when the other player steals, will always depend on considerations outside of the rules of the game. Ibrahim himself explained to Radiolab why he himself would prefer to steal, if he knew that Nick were stealing:
“I was always gonna steal … the reason being, if I split and the other guy steals, I get nothing. I’d rather both of us walk away with nothing, than someone embarrass me [by stealing when I split].”
For Ibrahim, the fear of embarrassment “broke the tie” in favor of stealing, should he believe that the other player was planning to steal. For others, undoubtedly, other emotional considerations such as revenge or envy also might break the tie in favor of stealing. But one could just as well imagine that, if some kind-hearted person were playing the game, he would genuinely prefer for the other person to get the money and hence share if he believed the other player planned to steal.
The beauty of Nick’s idea was to change Ibrahim’s payoff in the outcome “Nick Steals + Ibrahim Splits”, to give Ibrahim at least some hope that he would walk away with some money if he chooses “split” … whereas Ibrahim is SURE to get nothing if he chooses “steal”. It’s important to note here that Ibrahim doesn’t necessarily even need to believe Nick for this tactic to work. As long as there is ANY chance that Nick is telling the truth, and as long as Ibrahim cares more about the prospect of getting the money than the risk of being embarrassed, then Nick has now “broken the tie” for Ibrahim in favor of splitting the money.
This example illustrates the importance, in game-theory analysis, of paying close attention to “payoff indifferences,” situations in which players’ primary motivators (e.g. money in Golden Balls) don’t dictate their incentives. In these situations, secondary considerations (e.g. embarrassment for Ibrahim) will determine player strategy. For those looking to change games for the better, these situations are especially easy picking, as even a small change in the system (e.g. a chance that Nick is telling the truth and will share the money) is enough to dramatically change the natural outcome of the game.
- “NFL Coaches Are Too Chicken for Their Own Good” by David McAdams, February 2nd 2014 [Super Bowl Sunday], TIME.com.
On Super Bowl Sunday — yes, I’m very late posting this — I wrote a piece for Time.com on the game theory of football, in particular, on the games that coaches play. Here’s a tidbit:
“Even as many NFL coaches have opened their playbooks to new ways of moving the ball, few have embraced the more aggressive style of play that Romer showed would allow them to win more games. Why not? One possible answer has to do with another sort of game that coaches play, off the field.”
The New England Journal of Medicine recently featured an article (“Preserving Antibiotics — Rationally”) by economists Aidan Hollis and Ziana Ahmed, arguing for an antibiotic tax to discourage inappropriate use of antibiotics in livestock. Hollis and Ahmed’s argument boils down to an observation that ranchers’ use of antibiotics imposes a negative externality on the rest of society by increasing the risk of emergent antibiotic resistance, and that a tax can help by forcing farmers to internalize that cost.
Taxing antibiotics would indeed reduce ranchers’ antibiotic use, but whether this slows or hastens the rise of antibiotic resistance depends on whether the negative externalities that Hollis and Ahmed identify outweigh the positive externalities of antibiotic use that are entirely missing from their analysis.
To see the point, consider the problem of a rancher with a cow, sick with some drug-susceptible disease, that he would like to sell to another ranch. Eradicating the disease requires a two-week regimen of antibiotics, while temporarily suppressing its symptoms only requires a one-week regimen. Public policies that increase the cost of antibiotics will tend to push this rancher toward the one-week regimen, increasing the danger of antibiotic resistance even as fewer antibiotics are prescribed.
The problem here is that we want to encourage ranchers to use antibiotics appropriately on sick animals — society enjoys a positive externality in that case, as disease spreads less widely and appropriate antibiotic use minimizes the risk of rising resistance — while also discouraging them from using antibiotics inappropriately for growth-promotion. Consequently, society would actually benefit most if we were to subsidize some uses of antibiotics while taxing others.