WhatsApp Messages and “Confirmation Bias” on Turbo (Column 526)

A few days ago we received in our neighborhood WhatsApp group in Lod a message with a video and, beneath it, a caption:

In Lod: Two Arabs pummel a Jew because he came to exchange details with them after an accident

You can imagine how the discussion went after this message. But for those who cannot, I’ll devote this column to a guided account of it. I decided to present most of the discussion here with added explanations and extended notes, because in my view it is highly typical; it raised claims that come up in many similar debates each of you has surely encountered. I hope this annotated description helps the discerning reader deal with similar arguments and avoid the common pitfalls in them.

Background to the discussion

At first glance, the caption underneath simply describes the facts in the video. For those who don’t immediately catch it, in our context (especially in Lod, a mixed city that is hyper-sensitive to nationalist violence), this message says that once again we can see with our own eyes how Arabs habitually use violence against Jews in Lod and beyond. This video is taken as yet another representation and illustration—if anyone even needs one—of the general phenomenon of Arab violence toward Jews. No wonder this message landed on fertile ground among the group’s members and “proved” to all of us yet again what we always knew: Arabs are violent and, in particular, they employ nationalist violence against Jews.

Well, when I read the message and the replies to it, it seemed to me there were some people living on the moon. Let me remind those who have just now returned from a backpacking trip in the African savannah that in recent weeks, every single day we are informed of violent incidents on the roads (I’m not sure what caused this awakening). Following a dangerous pass, an accident, or even a provocative hand gesture, people beat each other with helmets, knives, clubs, brass knuckles, and for those without special equipment—just with their hands. So it seemed odd to me that someone would see a video like this as a demonstration of the violence of any particular group—Arabs, Poles, women, adolescents, or people 1.76 meters tall and under. To my eyes it’s just another video joining all the recent reports of violence, and from a broad view it doesn’t at all appear to characterize a particular population.

We could have taken a different video out of dozens on this subject and “proven” whatever we wanted: that motorcyclists, 46-year-olds, Belgians, or any other random category are prone to violence. Note that in the same article from which the report on this incident was taken, there’s mention alongside it of another similar incident involving a resident of Bnei Brak (I assume he is Jewish, by common criteria). For some reason, our WhatsApp post did not call our attention to the well-known violence of Bnei Brak residents.

The start of the discussion

So you won’t be surprised that yours truly replied to the above message as follows:

As is well known, Jews never do such things. All the news from the last few days must have come from the moon.

By the way, the video really doesn’t make clear what exactly happened there, and most of the blows were after the Jew also arrived and did things.

In short, there are stereotypes here, and the interpretation is far from what the caption implies.

I’ll add that regarding the other details as well, the video itself doesn’t make the picture clear to me. From watching it I got the impression the stick actually came out from the “attacked” party. I also don’t know where the information came from that they were Arabs (nor how the writer knew they were residents of Lod). But even if all that is entirely correct, my remarks above still stand.

The folks there rose up against the defense that a “leftist traitor” like me was allegedly providing our violent enemies. Here’s one example of a response:

True, there have been incidents lately, but to say these are just “stereotypes” is, in my view, a bit naïve. We’ve experienced enough events here in recent years to see that the Arabs in Lod are not good friends to Jews…

And you can also see whom Lod’s Arabs voted for in the last elections…

This is, of course, a logical fallacy, since even if we experienced that Arabs in Lod are violent or are not good friends to Jews, this video adds no information or support to that claim. If that’s what you think, fine, but even if your claim is true, not every piece of “evidence” you bring to it proves or strengthens it. A flawed proof of the Pythagorean theorem doesn’t become good just because the theorem it “proves” is true.

Alternatively, consider the following message:

Only Ben-Gvir will deal with these terrorists. Can’t you see they’re terrorists? Why judge them favorably?

Well then, even if they are indeed terrorists, you simply don’t see that in the video. You certainly see that they are violent, but that’s true of all the other stars of this week’s videos, regardless of their national, gender, religious, or sectoral identity.

On the obligation to judge favorably

Regarding the last bit—there being no obligation or license to judge Arabs (or terrorists?) favorably—I added the following reply:

If you look at the commentators on the Mishnah in Avot (Rabbenu Yonah and Maimonides) regarding judging favorably, you’ll understand this isn’t a mitzvah between man and his fellow but guidance for how to correctly understand reality.

My point was that the commentators explain it differently from how it’s commonly taken. They write that we are commanded to judge someone favorably only when reason indicates—or at least allows—such judgment. It’s not a categorical obligation toward our fellow (and thus it does not depend on whether he “acts as your people act”); rather, it is instruction on how to correctly judge reality.

As I have explained elsewhere (see my article on Occam’s razor, in Column 30 and at greater length in Column 440), when it’s a righteous person, the obligation to judge him favorably is not a moral duty toward him; rather, it is the more reasonable interpretation of reality. For example, if I see a well-known righteous man chasing a betrothed young woman in the public domain with a raised knife in his hand, it’s more reasonable to interpret that he intends to return to her the knife she forgot at his place in the kitchen than to interpret that he intends to rape her under threat. By contrast, with a wicked person that interpretation is less reasonable; therefore there really is no obligation to interpret it that way—not because we have no moral duties toward him since he’s wicked, but because that is not a reasonable interpretation of reality. Thus, the obligation to judge favorably isn’t a command toward one’s fellow but guidance on how to interpret reality. Therefore it applies to every person of any origin or sector, whether or not he “acts as your people act.”

You might say: It’s well known that Arabs are wicked, violent, and Jew-hating; therefore, for them, such an interpretation is the more reasonable one. Seemingly, from Avot the above speaker was right. Well—no. Even if that claim about Arabs is true as of today and based on solid facts, then perhaps that truly would be the correct attitude to the incident, and indeed we should interpret it as typical Arab violence against a Jew. But that interpretation rests on the premise (true, for the sake of discussion) that they are violent and antisemitic. If so, there’s no point or logic in bringing this video as evidence for the claim that Arabs are violent and antisemitic, since our interpretation of the video is based on that very claim.

What is this like? Someone “proves” that Belgians are violent by means of a video showing a violent person. He assumes that this person is Belgian because in his view (or based on prior knowledge) Belgians are violent, and now he uses this video to prove the thesis that Belgians are violent (since, according to his justified interpretation, we’re dealing with a Belgian). Again, even if it’s entirely true that Belgians are violent and we have solid knowledge to support it, such a video still cannot strengthen the thesis in any way. Moreover, even when we see a video in which someone we know to be Belgian behaves violently toward another person—if at the same time there are no fewer videos showing people of all nations behaving with identical violence—then too one cannot bring even a shred of evidence from such a video for Belgians’ notorious (and factually true, for the sake of discussion) violence. As stated, even if an argument is offered as proof of a true claim, that does not mean the argument is a valid proof of it.

A picture is worth a thousand words

Later in that message I wrote the following:

Anyone seeking reinforcement for his preconceptions will find it in anything that happens. A religious-Zionist learns from the Holocaust that one must be a Zionist, and a Haredi says the Holocaust happened because of Zionism. And so with every event. To “prove” from this case that Arabs are violent is a joke—especially in a week when many Jews are beating each other senseless on the road. Indeed, there are terrorists, and one cannot deny that there is nationalist violence, but not everything that happens is nationalist violence. This is just superficial thinking that strengthens the already strong and fixes our conceptions instead of weighing them on their merits. That way we’ll always come out righteous—but not very wise.

This is a chronic problem when we interpret texts or situations. Each of us interprets them in the direction that seems right to him and thereby becomes more entrenched in his position from everything he reads or sees. Thus the left will explain to you that the video proves Arabs are frustrated because we’re screwing them over and the lesson is that we must improve; and the right will explain that the video proves all Arabs are terrorists and we should kill them when they’re young. And of course, everyone as one will declare that this video is a picture worth a thousand words. Indeed, the persuasive power of a picture or video is like a thousand words, and that is precisely the danger in drawing conclusions from images or videos.

I remember that during the First Gulf War there were images broadcast of cormorants floating in oil-filled water and dying en masse (see, for example, a somewhat demagogic use of this here). We were all horrified, but I suddenly thought that at the same time countless people there were dying and losing their property and their lives, while we were focusing on the fate of the cormorants. A picture is indeed worth a thousand words—and that’s a pity. Images and videos are a first-rate misleading tool precisely because of their power. They neutralize our ability to approach an issue in a balanced manner, because they inflame emotion and automatically reinforce what we thought beforehand. They literally take us and our thinking captive. Sometimes it’s done deliberately, and then the disseminator of the picture or video is a demagogic propagandist; but in most cases it happens innocently. Still, in both cases our interpretation of such materials is flawed and biased.

Confirmation bias

In the continuation of that WhatsApp thread, a discussion developed with one of the participants.

The images are harsh. Sadly, violence has also become the lot of the Jewish public, but it’s also clear that we live alongside Arab society, where violence crosses boundaries and knows no limits. Most of the time it’s one against the other, and not infrequently it spills over on us. One of the rioters seems to have come out with a stick. I don’t have a stick in my car. But we have no right to complain about that, when these rioters have no problem carrying out mutual hits in front of a police station. In those days there was no king in Israel; each man did what was right in his own eyes.

This description sounded fairly accurate to me, but my claim was that its connection to the video is problematic. Even in the details there are issues, since we’ve seen quite a few Jews getting out of their cars with sticks or knives, even if neither I nor the writer has such sticks in our vehicles. I assume there are also quite a few Arabs who don’t have sticks in their cars. Again, irrespective of the truth of the claim, the video doesn’t constitute any corroboration of it.

At this stage I posted a link there to the Wikipedia article on “confirmation bias.” In brief, it’s a known phenomenon whereby a person sees in every phenomenon reinforcement for his own positions and tends to filter information according to his beliefs, giving greater weight to what supports his views and reducing the weight of—and distorting the interpretation of—materials (facts and arguments) that contradict them.

He replied:

I didn’t understand the claim. Just as the fact that a person suffers from paranoia doesn’t mean he’s never actually in danger, so too the fact that we tend to remember what fits us and our beliefs is not proof that we’re wrong. And no—the violence in Jewish society does not reach the level of violence in Arab society (though that’s no great source of pride).

I answered:

Obviously it’s not proof that you’re wrong, but if you’re paranoid you must be more wary that you may be mistaken when you think you’re being chased. That’s what confirmation bias implies. If you think Arab society is violent, that’s perfectly fine—and you’re probably right. But the video posted here is no evidence of it.

Is the problem inferring from a single case?

Here came a further refinement from my interlocutor:

Clearly there’s no proof here. There is never proof from a single case.

A typical and seemingly persuasive answer—but actually patently wrong. He thought my claim was that a single case can never prove a general claim. As is known, induction is risky and should be used with caution. And he argued that at the same time generalization is a cornerstone of our thinking. One case plus another amounts to a good generalization. In that sense, he said, the video does not prove Arab violence but is an illustration that, together with other examples, can corroborate that thesis. The video contributes its two cents to the overall picture.

Where is the mistake in his words? I answered briefly:

A single case can serve as a small piece of evidence that joins a larger account. Here there isn’t even that.

I meant to clarify that my claim was not that a single case proves nothing. That’s trivial and not worth debating. My claim is that this video does not give us even the minimal corroboration one gets from a single case. It changes nothing regarding the general thesis. Why? Because a single case can slightly corroborate a general thesis when it truly points in its direction. If there were one video of Arabs beating a Jew and there were not comparable videos about Jews, then there would be some corroboration—small, but still—for the general thesis about Arab violence. But when there are no fewer videos of Jewish violence, the standing of such a single video evaporates. Now it’s not a single example; rather, there’s no example here at all. In short, my claim was that this video neither adds nor subtracts anything in the discussion.

The entanglement effect: combining echo, confirmation bias, and the fake phenomenon

We’ve seen that WhatsApp messages of the kind I described serve as a trigger and reinforcement for existing positions due to confirmation bias. To sharpen the point, I should add two more important aspects that amplify the effect. The first is what I described in Column 335. We saw there that social networks echo back to each person the positions, information, and arguments he cherishes. Algorithms ensure that we are exposed primarily to information and articles that fit our positions; thus our positions receive constant amplification, and our confidence in them grows ever stronger—without our noticing that we live in the matrix. What once happened on a small, manual, human scale in the print press now happens automatically and on a large scale on the internet and especially on social networks.

The second aspect is the fake phenomenon. The web is full of incorrect “facts,” fabricated images and videos, and many viewers (even skilled ones) fall into the trap because of the echo and confirmation bias. These two effects deepen error even in the face of genuine material, and they certainly intensify the effect of fake materials.

The conclusion is that we get a cumulative impact woven from three phenomena: confirmation bias, the echo effect, and fakes:

1. Confirmation bias leads us to see in every picture, video, or text reinforcement for our current positions—rightly or wrongly. This is a distortion in interpretation even when we are presented with a balanced and accurate picture.
2. Now add echo. Filtering by social networks (sometimes done automatically by algorithms like those of Google and Facebook, sometimes manually—say, by members of our WhatsApp group) causes us to be presented mainly with material that reinforces our positions, while material that contradicts them is filtered out. The picture before us is not balanced but agenda-driven and biased.
3. All this operates on genuine materials, and can operate far more severely on the basis of fake materials. Here we are usually dealing with intentional and engineered biases.

This triple entanglement has a destructive effect on our ability to exercise critical thinking and reconsider our positions. In fact, it doesn’t even allow us to converse and debate and listen to one another. We turn ourselves, with our own hands, into zombies, each trapped in his own monad (or narrative), and thus postmodern narrativism realizes itself ex nihilo. A foolish, baseless theory that claims there is no truth and no justice turns, before our bewildered eyes, into a frustrating reality to which we all find ourselves enslaved.

Note that the more convinced we are that absolute truth is on our side, the more it proves—or actually creates—the opposite situation: that there is no absolute truth, only a collection of narratives. My war against the blind, absolutist righteousness of our WhatsApp, Facebook, and Google bubbles aims to save truth and justice, not attack them. It is a war against narrativity, not for it. It’s important to understand that, ironically, it is precisely our sense of absolutism and our unwillingness to hear and weigh other narratives that brings about narrativity.

An unavoidable note on Aggadah, Hasidut, and Tanakh

Incidentally, this is exactly what I claim happens in the study of aggadot (see Column 452 and elsewhere), Hasidic teachings (see Columns 104–105), and Tanakh (see Columns 134–135, 361, and more). In those columns I clarified why I am not inclined to engage with these genres, and here I’ll sharpen and reinforce my claim. I argued there that, as a rule, engaging in these areas does not constitute study (and except for Tanakh, in my opinion they cannot even fall under the heading of “Torah”), and therefore there is little value in engaging in them. Here I add that such engagement even has a harmful side, since it usually reinforces what we thought from the outset. Each person interprets the aggadah, the Hasidic teaching, or the biblical portion according to his beliefs and values, and thereby fixes and perpetuates errors and locks us into the bubbles within which we live. Such “study” makes it very difficult to examine our existing positions in a balanced way, and certainly to change them when necessary.

A side note: How the media reports problematic phenomena

There is a policy and journalistic code of ethics (not always observed) according to which the media should not report the origin, sector, or gender of those involved in problematic activity (such as traffic accidents, sexual harassment, crime, violence, and the like). This always irritated me, since I thought the public deserves information about what is happening, and that the concealment—which distorts the picture—stems from political correctness, which I detest.

But in light of what I’ve written here, it seems the issue deserves reconsideration. Such full reporting, though proper in itself, may trip many people up through the entanglement effect I described. Not coincidentally, the caption in our WhatsApp spoke of “two Arabs who beat a (naïve) Jew,” while in the media the description was more sterile and neutral. Not Arabs, not a Jew, and not “beat to a pulp.” It was violence following the exchange of details after an accident. That is a purely factual description; admittedly it lacks information, but at least it isn’t misleading.

There would be room to qualify this policy and exempt cases in which sectoral/gender information is relevant to the events, or when there is a clear public interest in knowing about the involvement of a specific sector in the activities in question. But as we have seen, the decision itself is problematic: Is the knowledge that Arabs were involved relevant in our case? The members of our neighborhood WhatsApp group would say yes. I, on the other hand, think not. So who decides? The journalist? That opens the door to biases stemming from the journalist’s views. Therefore I’m beginning to see the logic in not deviating from this policy even in cases where there could be justification to deviate.

If we could improve our interpretation of the materials placed before us and our ability to treat them objectively, then there would be no justification for concealing information. Our lack of skill and our interpretive failures due to the entanglement effect, sadly, provide some justification for this annoying and paternalistic policy. There you have another reason to improve in this matter.