Recently I attended the OOPSLA 2021 conference. While I haven’t barely learnt anything from the conference itself, it’s indeed a break from my routine life, and I have probably met more people than the sum of the past few months. So as a result, I had some random thoughts and reflections, which I decide to take notes here before they disappear.
Methodology of Decision Making
It is well known that one should not judge a decision based on its outcome, because we cannot have full awareness of the world (so there are information that we cannot know beforehand), and the world itself also contains many random factors (so we cannot predict the future even with complete information). Therefore, an undesirable outcome does not imply the original decision is wrong or improvable.
However, humans are born irrational. Many never understand the above argument at all. But even within the people who can recognize it, I have seen (on both myself and others) many cognitive pitfalls when applying the argument.
Pitfall #1. Only apply when things go wrong. Ego makes people believe in their own decisions. So when things work out, it’s easy to overlook the possibility that the success is only coincidental and the original decision is unjustifiable.
As an extreme example, winning a lottery does not justify the decision of spending large amounts of money buying lotteries (which has a negative expected gain).
Pitfall #2. Overlooking the information one can know. Similarly, due to people’s ego, when something fails, it’s easy to use the argument as an excuse to deny responsibility, not realizing that the original decision could have been improved with more investigation and reasoning.
The most obvious examples are decisions made through overconfidence and negligence.
Pitfall #3. Overlooking the information one cannot know. As stated earlier, one cannot have complete information of the world, so that “there are information that one cannot know” is also an information that must be considered in decision making.
Examples of this pitfall are “perfect plans” that are designed without backups and leaves no buffer on accidents and errors.
So in short, one should not judge a decision based on its outcome, be it desirable or not; one should judge a decision based on the justifiability of how the decision is reached.
(Interestingly, everything stated above can also be observed in Japanese Mahjong)
Methodology of Becoming Productive Researchers
I am aware that all my research ideas have been produced by pure luck. If one had rewinded time, I’m very doubtful if I could come up with the same ideas again. And I feel it clueless to figure out the “next idea” like some of the PhD students who are more “on the right track” could easily do. So I have been curious how the professors can generate an endless stream of ideas for papers. I happened to have discussed this topic with two professors, so for future reference, I take notes here based on my memory.
Q1: (The context here is theoretical computer science.) The difficulty of figuring out a proof is probably exponential to the number of the steps in the proof. So how can you and your group produce so many long (>50 pages) papers every year?
A1 (Richard): We are not going for particular problems. Instead, we have a set of powerful math tools as building blocks, and we just build things up by gluing the blocks together, without a particular goal in mind. If at some time we found that the thing we built solved some problem, we have a paper. It’s like a naval battle: you don’t search for and destroy a particular ship (problem) on the sea. You patrol the sea and destroy any ship spotted along the way.
Q2: But I assume you still have some intuition on what might work and what might not. What is the intuition that guides you and how did you get this intuition?
A2 (Richard): I don’t know. It’s like the first man who figured out they can put guns on ships.
Q3: (The context here is PL/Compilers, and my advisor primarily works on sparse tensor algebra.) I asked my advisor how he could always have the next paper idea to work on.
A3 (Fred): There are many solved problems in the dense algebra domain, but little is known about the sparse algebra counterparts. TACO is a framework for solving problems in the sparse algebra domain, so it opens up a sequence of works by porting the solved problems in dense algebra to sparse algebra.
What Prevents Constructive and Rational Discussion?
The motivation of this part is the (still ongoing!) Chinese Internet shitshow centered around a consipracy theory that the Grand Final match of Dota2 TI10 is fixed.
I have never held any expectation on the rationality of the mass public, but I’m still astonished by that a conspiracy theory without even a self-coherent story can get dominance in the Chinese Dota2 community.
Though it might be my illusion, but I do feel the Internet discussions I see on contemporary matters have been getting increasingly polarized / emotion-driven, and decreasingly constructive / helpful in the past years. Is the mass public becoming more irrational? I don’t know. But after thinking for a bit, I do feel there are a few contributing factors.
- It takes much more words and time to refute a conspiracy theory or to post some serious discussion, than to propose a conspiracy theory or to post some trashtalk.
- The bandwidth of the Internet has greatly increased, but the bandwidth of the useful information being carried has actually decreased. On one hand, blogs are replaced by tweets, texts are replaced by pictures/videos, so the mass public has been trained to only read short messages, not serious discussions. On the other hand, mobile phones, which are not even designed to type efficiently, have surpassed the market occupation rate of PCs long ago, so it is also harder for the mass public to publish anything other than short messages. So the mass public has been trained to only read 140 characters and post 140 characters, not any serious discussions.
- By Pareto principle, 80% of the voice in a community comes from 20% of the people. And the people who feel most compelled to speak out are usually the people holding the most extreme opinions. But under the current shape of the Internet, where serious discussions are unfavored, the megaphone is handed over to the most irrational ones, not the most rational ones.
- The ranking mechanism, where contents are ranked by user votes and shown to users by rank, served as another amplifier.
- But what about the POLs? Will they send out rational messages and lead the public opinions to the rational side? Unfortunately, at least for the Chinese Internet’s status quo, where most of the POLs are commercialized, the answer is negative. The POLs do not care about anything but making more money, which come from public exposure and supporters. So they have no motivation to argue against the trend at all. In fact, many POLs are known to intentionally start flamewars or spread falsehood messages to gain exposure.
- Another interesting factor is the bots. It might be surprising, but a CMU research showed that the majority of the COVID falsehood messages on Twitter are spreaded by bots. And while I haven’t seen an academic research for the Chinese Internet, it’s undeniable that there are many keyword-based bots for various purposes (lottery, advertisement, promotion, PR manipulation, etc). It’s not surprising that there are similar falsehood message bots as well.
But what exactly went wrong? And how this might be fixed? Honestly I don’t know.