“A few months ago, I told the American people I did not trade arms for hostages. My heart and my best intentions still tell me that’s true, but the facts and the evidence tell me it is not.”
– President Ronald Reagan (March 1987)
“It is not a lie if you believe it.”
– George Constanza character on Seinfeld (1990)
“You take the blue pill…you wake up in your bed and believe whatever you want to believe.”
– Morpheus Character, The Matrix (1999)
Blying – (noun) the act of communicating beliefs unlikely to be true, selected for benefit and believed to a high degree by the communicator
In 2004, I observed an interesting fact. In the Bush-Kerry debate, the score of verifiably untrue statements was (Drum 2004):
- Bush # verifiably untrue statements = 18
- Kerry # verifiably untrue statements = 11.
Many people recoil in horror at such information. “Politicians are all liars,” they might mutter and feel disgust. Yet, it is widely reported that President G. W. Bush was sincere in many of his beliefs, even while many were verifiably false.
Here, I argue that this score is part of an interesting phenomenon. Would you or I make 11 verifiably untrue statements in a debate? Would we have any chance to compete at such a high political level? Is it possible that false statements offer a key advantage at such a level?
In 2006, I endeavored (somewhat halfheartedly) to model this phenomenon. I also invented the concept of blying and gave it a name googling options. At that time, blying was listed as a sailing term. I presented the concepts at a university departmental seminar. Some tend to think of such seminars as obscure in importance with mostly unassuming foreign graduate students attending. I asked the attendees to use the word and spread its use. Amazingly, the word has gained in popularity with celebrities such as Rachel Maddow and Nina Dobrev using it in public.
I never sought to publish my ideas relating to blying but here are some key slides from my old presentation.
Consider the set, S, of things that politicians can say that voters might consider to be believable or at least not ridiculous as indicated in Figure 1. Fortunately, not everything is in the set S. Saying that people should drink disinfectants to address COVID-19 was recently found not to be in the set. Plagiarized words may not be in the set S. Also consider the set of things that politicians might say, and experts would agree that they are plausibly true, R (right-hand-side of Figure 1).
Figure 1. The sets S and R relating to political discourse.
Given the competitive nature of politics, it is unrealistic to imagine that politicians will restrict themselves to set R. In fact, staying away from R might confer benefits as ivory tower intellectuals will tend to be on the other side. Yet, in this information age of hopeful enlightenment, it might be possible (theoretically) to shrink the set S until it fits inside set R. Then, even competitive politicians would find themselves in agreement with at least some respectable experts. This is perhaps the main goal of factSpread.
Perhaps the quintessential illustration of blying relates to supply-side economics. Clearly, at some level taxation rates will become counterproductive. For example, if you take 99.9% of my last dollar earned, I might well take more vacations and earn less. Yet, evidence that this phenomenon has applied in the context of actual US tax rates and economic growth is limited at best. This evidence is the subject of another bulletin post. Here, I use the example for my half-hearted modeling.
In 2006, I imagined the political process as a two-player competitive game influence by Filar and Vrieze (1999), who wrote on “Competitive Markov Decision Processes”. It should be noted that: (a) Markov Decision Processes have formed the core for the hot subject of “Reinforcement Learning” within machine learning which is the subject of many books including my own forthcoming textbook with Enhao Liu and (b) game-theoretic interactions are complicated and I am only now becoming able to actually use such advanced methods in modeling.
Player 1 is 95% of the US public which could control politics in the US if it wanted to. Player 2 is the 5% who have limited direct influence but can control things if only through their influence of Player 1. (This was before the “1%” concept became popular.)
The key idea that I introduced was that Player 2 influences the system only through the selection of beliefs. Further, these beliefs drag the beliefs of Player 1 away from its original beliefs. Note that in all or most modeling activities, beliefs are generally regarded not as controllable or decision variables. Thinking of beliefs as controllable was modestly innovative.
Back then, the states, defined by Gross Domestic Product levels, were sketched in the big picture system as shown in Figure 2 . (The economy is now considerably larger, projected to exceed $20 trillion by the end of the year despite the contraction caused by the pandemic)
Figure 2. A super-simple model of the economic state of the US economy.
In Markov Decision Processes and their generalization Partially Observable Markov Decision Processes (POMDP), beliefs can be represented in part by transition and reward matrices. Figure 3 sketches some possible beliefs about transition probabilities or rewards. Perhaps the default belief for the 95% is that the choice of tax rate (level 1 or level 2) has minimal if any effect on the economic trajectory for relevant levels.
Figure 3. The default actions for Player 1 (the 95%) that tax rates effect revenues but not the overall GDP.
The objective of Player 2 is to maximize its expected profits. It can convince Player 1 to set the rate and then reap rewards personally. All it needs to do is “believe whatever they want to believe” and convince Player 1 through the convex combination in Figure 4.
Figure 4. The optimization “objective function” with the constraint which relates to blying and influence.
Player 1 on its own would simply keep the top tax rate at 70% in all cases since it would keep high levels for all system states. Figure 5 shows the single-player solution and backward recursion.
Figure 5. The optimal policy and the usual dynamic programming recursion.
Yet, with two players and blying, we wind up with a top tax rate as low as the set S will permit, e.g., 37%. In this solution, we wind up with a self-brainwashed Player 2 and a partially brainwashed Player 1 and a large governmental debt.
Why would people (or even robots) deceive themselves and say verifiably untrue statements? Part of the reason is to convince others and, therefore, receive benefits. Our faces can give us away if we are merely lying. Deeply held beliefs are more likely to influence.
In conclusion, the phenomenon of blying will continue to greatly influence politics in democracies and, to a lesser extent, other human activities. The key implication from the related discussion is the importance of reducing the set S – set of falsehoods people consider believable – by educating the public to tell the difference. This is the primary mission of factSpread.
- Filar, J., & Vrieze, K. (2012). Competitive Markov decision processes. Springer Science & Business Media.