Tumbling Chimp

A country can’t really qualify as a democracy without civilian control of its own security forces, but the logic that makes that statement true also makes civilian control hard to achieve, as events in Burkina Faso are currently reminding us. The essential principle of democracy is popular sovereignty. Civilian control is fundamental to democracy because popular sovereignty requires that elected…

Economic performance is a key component of most election forecasts. When fitting models, however, most forecasters unwittingly assume that the actual state of the economy, a state best estimated by the multiple periodic revisions to official macroeconomic statistics, drives voter behavior. The difference in macroeconomic estimates between revised and original data vintages can be substantial, commonly over 100% (two-fold) for economic growth estimates, making the choice of which data release to use important for the predictive validity of a model. We systematically compare the predictions of four forecasting models for numerous US presidential elections using real-time and vintage data. We find that newer data are not better data for election forecasting: forecasting error increases with data revisions. This result suggests that voter perceptions of economic growth are influenced more by media reports about the economy, which are based on initial economic estimates, than by the actual state of the economy.

In attempting to understand [social] movement effects on media, movement scholars have sometimes, but certainly not always, conceptualized media-movement interactions within what I would call the “bias model.” The idea behind the bias model is that media attention and framing are subject to numerous organizational, cultural, political, and institutional selection processes which filter movement messages and events to determine which will receive coverage and how they will be framed. That is, some population of movement events and messages exist in the world, and are distorted in news media representations through differential media selection and interpretation. Within the bias model, the media nicely fits the analogy of a movement target—a wholly separate entity at which a movement takes aim. While we’ve learned a lot from the bias model, it is incomplete, and often misleading. The media-movement relationship is endogenous.

Across practices, across cultures, and throughout historical periods, when people support and engage in violence, their primary motivations are moral. By ‘moral’, I mean that people are violent because they feel they must be; because they feel that their violence is obligatory. They know that they are harming fully human beings. Nonetheless, they believe they should. Violence does not stem from a psychopathic lack of morality. Quite the reverse: it comes from the exercise of perceived moral rights and obligations.

Often when we study social problems, there is an almost irresistible temptation to study what we would like the cause of those problems to be (for whatever reason), to the neglect of the actual causes. When this goes uncorrected, you can get the phenomenon of “politically correct” explanations for various social problems – where there’s no hard evidence that A actually causes B, but where people, for one reason or another, think that A ought to be the explanation for B. This can lead to a situation in which denying that A is the cause of B becomes morally stigmatized, and so people affirm the connection primarily because they feel obliged to, not because they’ve been persuaded by any evidence.

Indeed. I wrote something on this topic a while back, too.

To help discover how well a self-selecting crowd with no money at stake can predict the winner of sporting events like this year’s Tour de France, please cast some (or many) votes in the pairwise wiki survey I’ve created at this link:

2015 Tour de France Winner Wiki Survey

A pairwise wiki survey involves a single question with numerous possible responses—here, rider names—on which participants vote in pairs, one randomly-selected pair at a time, as many times as they like in as many sessions as they like. A Bayesian hierarchical model converts those pairwise votes into the aggregate ratings that can be treated, in this case, as a predicted finishing order.

So far, I have used these surveys to generate predictions of NFL games and seasons, mass-killing onsets, and coup attempts. With this year’s TdF fast approaching, I thought it would be fun to apply the method to my other favorite sport, cycling, where it’s arguably a more natural fit because rank ordering is what we care about.

Data in the wild: a histogram of Yahtzee scores from old score sheets found in the box in our house (n = 28) with an overlaid gamma density curve.

Here, gamma (shape = mean^2/var, scale = var/mean) fit better than normal.

From inspecting the score sheets, I can tell that this sample was shaped by a selection process. A number of sheets had unfinished games, or games that were finished but not tallied. I suspect these are from times when one of my kids was clearly losing. I tallied and included the latter, but I couldn’t include the former.

This exercise got me wondering about the limits of this distribution. I found a Yahtzee fan page that bounds scores at 5 and 1575. That page also notes that “in normal play, where an experienced player is trying to get a high score,” scores rarely fall below 150. This is why the gamma distribution fits better; although very low scores are possible, they can only occur if you do what amounts to playing a different game --- deliberately score low --- and, in normal game play, the lower bound is much closer to the mean than the upper bound. That page also notes that the best score without a yahtzee is approximately 270, and “most times a good score is in the 250 to 260 zone, maybe pushing to 270.” You can see those stylized facts in the data from our house.

Finally, I also found a page that offers online play and collects and displays data about the scores of those games from users who register. Below, you can see the distribution of the much larger sample collected there. In that histogram, we see the same bump in the high 300s and low 400s that’s present in the sample from my house. I initially figured that bump was noise, but it actually comes from the 100-point bonus for a second yahtzee. With so many observations, we can even see another but still smaller hump in the high 400s from bonuses for third yahtzees. Those bonuses produce a multimodal distribution; it’s just hard to see the additional modes because multiple-yahtzee games are so rare.

I went to the 1 o'clock showing of Ex Machina at the local multiplex yesterday. There were eight people in the theater including me, all of us men. Two guys who arrived a few minutes after the show had started sat two seats away from me, and the closer one made a small bag of popcorn last through the entire film, slow-eating it one piece at a time. Ex Machina is a quiet movie, with almost no background music and not much ambient sound. If you know me well, you'll know that I almost chewed my own arm off, going crazy at the sound of him intermittently rustling that bag and eating that corn.

In spite of that glitch in my personal matrix, I enjoyed the movie very much. I happen to be reading Steven Pinker’s Better Angels of Our Nature right now, and the two works dovetailed nicely. Among the many ideas swirling through Ex Machina, the one that struck me strongest --- because of the connection to Pinker’s work, I think --- was the reminder that intelligence and morality are very different things. The two have co-evolved in our species, but our moral impulses stem from our social nature, not our intellect as such.

We might --- might --- be able to build machines that think autonomously and accurately read our social and emotional cues. If we do, though, we should not expect senses of compassion, fairness, and shame to follow. Instead, we ought to worry not only about being out-thunk by our creations, but also about being outfoxed by entities that can reason through how to manipulate our moral impulses without caring how we feel about the consequences. Instead of worrying so much about creating Terminators, perhaps we should worry more about creating psychopaths.

I also appreciated Garland’s variation on the theme of the allegory of the cave. In Plato’s version, Socrates believes that the prisoner who first ascends and becomes enlightened would want to return to free his former cellmates, even though he has to beware of the risk to himself from those cellmates’ fear and misunderstanding. Both curiosity and pity are essential to the allegory’s arc.