Name: 模擬綠鬍子利他行為！ (Simulating Green Beard Altruism)
Uploaded: 2025-11-19T00:52:31.000Z
Duration: 13 min 23 s

轉承詞

Thanks to Brilliant for supporting this video.

感謝 Brilliant 對本影片的支持。

If evolution is all about the survival and reproduction of the fittest individuals, then why are creatures sometimes nice to each other?

如果進化就是適者生存和繁殖，那為什麼有些生物有時候會對彼此友善呢？

In this video, we're going to build a simulation to start looking at the biology of altruism.

在這部影片裡，我們要來建立一個模擬程式，開始探討利他主義的生物學。

Certain genes will code for certain types of behavior rules, and

某些基因會編碼特定的行為規則，

we'll see which kinds work from a gene's perspective.

然後我們會從基因的視角來看看哪些規則有效。

那我們的模擬程式長什麼樣子呢？

Each day, blobs will go out to the forest to eat some fruit.

每天，點點們都會去森林裡吃點水果。

Each tree has enough fruit to feed two blobs, but some of the trees have predators in them.

每棵樹都有足夠的水果餵飽兩個點點，但有些樹裡有掠食者。

If the blobs are unlucky enough to visit a tree with a predator, they'll get eaten.

如果點點們不幸遇到有掠食者的樹，牠們就會被吃掉。

But if they don't get eaten, they'll head to one of the homes and reproduce, creating one or two new blobs in their place.

但如果牠們沒被吃掉，牠們就會回到其中一個家進行繁殖，產生一到兩個新的點點。

let's make this a bit bigger and let it run for a while to see how it goes.

我們讓它跑久一點，看看會發生什麼。

Alright, so this little ecosystem is mostly stable.

好的，所以這個小小的生態系統大部分是穩定的。

Now let's add the ability to be altruistic.

現在我們來加入利他行為的能力。

When the blobs go to a tree with a predator in it, one of them will notice the predator and it'll have two options for how to react.

當點點們走到有掠食者的樹時，其中一個會發現掠食者，牠有兩種反應方式。

The first option is to just run away, leaving the other blob to die.

第一種就是趕快逃跑，留下另一個點點等死。

我們把這種行為叫做懦弱行為。

The second option will be to make a bunch of noise and warn the other blob, but

第二種選擇是發出很大的聲音警告另一個點點，但是

in the process attract the predator's attention with a 50% chance of getting eaten afterward.

這樣會吸引掠食者的注意，之後有 50% 的機率被吃掉。

Since these blobs are risking themselves to help others, we'll call this the altruistic behavior.

既然這些點點冒險幫助了其他點點，我們就稱這種行為為利他行為。

The blob's behavior will be controlled by a gene which has two versions or alleles.

點點的行為會由一個基因控制，這個基因有兩個版本，也就是等位基因。

The cowardly allele will be yella and the altruism allele will be blue.

懦弱的等位基因是黃色，利他的等位基因是藍色。

Before we get to the simulation, it's worth saying that in real life, behavior is determined by a mind numbingly complex interaction between many genes and a creature's environment over time.

在我們開始模擬之前，值得一提的是，在真實生活中，行為是由許多基因與生物的環境隨時間發展所產生的令人麻木的複雜互動所決定的。

But this simple model is good enough to help us grasp some fundamental ideas.

但這個簡單的模型足以幫助我們掌握一些基本概念。

Okay, let's start with equal numbers of each allele and see what happens.

好的，我們先從兩種等位基因數量相等開始，看看會發生什麼事。

This graph shows the frequency of each allele over time.

這張圖顯示了每種等位基因的頻率隨時間的變化。

At the beginning, half the alleles are the blue altruism kind and stacked on top.

一開始，一半的等位基因是藍色的利他類型，疊加在上面。

Half the alleles are the yellow cowardly kind.

一半的等位基因是黃色的懦弱類型。

These values change over time, but they always add up to 1 or 100%.

這些數值會隨時間改變，但它們加起來總是等於 1 或 100%。

And speaking of the change, it looks like the altruism Allele goes extinct after just a few generations.

說到這個變化，看起來利他主義的等位基因在幾代後就滅絕了。

Maybe we made it a little bit too dangerous to be altruistic.

也許我們讓利他主義變得有點太危險了。

So let's lower the death chance for being altruistic from 50% to 10% to see if the altruism allele can do well on easy mode.

所以我們把利他主義的死亡機率從 50% 降到 10%，看看在簡單模式下利他主義的等位基因表現如何。

Okay, it took a bit longer, but the altruism allele still went extinct.

好的，雖然花了一點時間，但利他主義的等位基因還是滅絕了。

Now you might think, hey, wait a minute, we only tried this once.

現在你可能會想，嘿，等一下，我們只試了一次。

So now let's repeat 30 versions of the same situation and see how that looks predict ahead of time what you think will happen.

所以現在我們來重複 30 次相同的模擬，看看結果如何，並預測一下你認為會發生什麼。

Do you think the altruism allele will always go extinct?

你覺得利他主義的等位基因會一直滅絕嗎？

Or maybe the first run was a fluke and the altruism allele will actually increase in frequency more often than not.

或者也許第一次的結果只是個巧合，而利他主義的等位基因實際上會比不常增加頻率。

Our brains like to pretend we knew things all along.

我們的大腦喜歡假裝我們早就知道一切了。

So it's good learning and science practice to make a prediction before you see a result.

所以，在看到結果之前先做預測，這對學習和科學練習都很有幫助。

Okay, there's a lot going on in this graph.

Each thin black line is one run, and the thick black line is the average.

每一條細黑線代表一次運行，而粗黑線是平均值。

In some runs, the altruism allele lasted longer than others, but extinction seemed inevitable.

在某些運行中，利他主義的等位基因比其他運行持續得更久，但似乎都難逃滅絕。

Well, one did hang on for all 400 generations, but

嗯，有一個確實持續了全部 400 代，但是

based on the other runs, it seems like it's pretty doomed in the long run.

根據其他的運行結果來看，長遠來看它似乎是註定失敗的。

So why did this version of altruism fail?

那麼，為什麼這個版本的利他主義會失敗呢？

Well, this is a computer simulation, so we have a perfect record for everything that happened.

嗯，這是一個電腦模擬，所以我們對所有發生的事情都有完整的紀錄。

There were 225,817 blobs saved by warning calls.

有 225,817 個點點因為警告電話而獲救。

85,781 had the altruism allele, and the other 140,036 had the cowardly allele.

其中 85,781 個帶有利他主義等位基因，另外 140,036 個帶有懦弱等位基因。

And 24,000, 151 of the altruistic blobs died after giving the warning.

而在那些給出警告的利他點點中，有 24,000 個（151 個）死亡了。

So overall, the altruistic allele saved a Net total of 61,630 copies of the altruism allele.

所以總的來說，利他主義等位基因淨拯救了 61,630 個利他主義等位基因的複本。

But since the total population size is limited by the predators, the total number saved isn't really what matters.

但由於總人口數受到掠食者的限制，真正重要的不是總數獲救的數量。

More blobs being saved just results in more blobs getting eaten later on.

更多點點獲救，只會導致之後有更多點點被吃掉。

What does matter, though, is the competition between alleles.

然而，真正重要的是等位基因之間的競爭。

A successful allele for altruism would need to find a way to help itself more than it helps its competitor.

一個成功的利他主義等位基因需要找到一種方法，讓自己比競爭對手獲益更多。

Well, one way could be to let the altruism alleles see each other so they know who to help.

嗯，一種方法是讓利他主義等位基因互相看見，這樣牠們就知道要幫助誰。

But because DNA stays inside the blobs and it doesn't have eyes, we need some detectable feature to go with the altruism.

但因為 DNA 留在點點裡面，而且牠們沒有眼睛，所以我們需要一個可偵測的特徵來搭配利他主義。

The classic example of this is a Greenbeard, and

這方面的經典例子就是綠鬍子，

that's a fun thing to put in the blob, so let's stick with that.

這是一個很有趣的東西可以放進點點裡，所以我們就用這個吧。

So instead of the allele for altruism from before, we have an allele for green beard altruism now.

所以，我們不再是之前的利他主義等位基因，而是綠鬍子利他主義等位基因了。

When a blob with a green beard notices a predator, it'll only warn the other blob if that blob also has a green beard.

當一個有綠鬍子的點點發現掠食者時，牠只會警告其他也有綠鬍子的點點。

So let's see whether this Greenbeard strategy works in a simulation.

所以，我們來看看這個綠鬍子策略在模擬中是否有效。

And let's go back to the 50% chance of getting eaten after giving a warning.

然後，我們回到發出警告後被吃掉的 50% 機率。

The green beard allele drove the cowardly allele to extinction, and decisively so.

綠鬍子等位基因讓懦弱等位基因滅絕了，而且是徹底滅絕。

But again, we should test it with many runs and with different settings.

不過，我們還是應該用很多次模擬和不同的設定來測試。

I ended up testing the 50% and a 90% chance of getting eaten for 30 runs each.

我最後測試了 50% 和 90% 的被吃掉機率，各跑了 30 次。

In each case, what do you think will happen?

在這種情況下，你覺得會發生什麼事？

At the 50% death chance, it seems like our first run wasn't a fluke.

在 50% 的死亡機率下，看來我們第一次的結果不是僥倖。

The Greenbeard allele does indeed have a massive advantage.

綠鬍子等位基因確實有巨大的優勢。

One of the runs got a pretty rough start, but

其中一次模擬的開頭有點坎坷，但是

it still ended up taking over before too long.

在 90% 的情況下，情況就比較混亂了。

Only 10% of the acts of altruism actually end up saving an extra blob.

只有 10% 的利他行為最後能拯救額外的點點。

So the random fluctuations are pretty dominant.

And even though the average shows an upward trend, the green beard allele does go extinct in several runs.

即使平均值顯示上升趨勢，綠鬍子等位基因在幾次模擬中還是會滅絕。

After seeing these, I was curious to try again.

看到這些之後，我很好奇想再試一次。

But starting with only a 10% frequency for the green beard allele once again, what do you predict?

但這次綠鬍子等位基因的起始頻率只有 10%，你預測會怎麼樣？

Will the graphs look the same, just shifted down to start at 10%?

圖表看起來會一樣，只是往下移到從 10% 開始嗎？

In the 50% case, there are a few interesting things going on.

在 50% 的情況下，有些有趣的現象正在發生。

After a while, we see that the average goes completely flat, because in every simulation, one allele or the other has gone extinct.

過了一陣子，我們發現平均值變得完全平坦，因為在每次模擬中，其中一個等位基因都會滅絕。

Of the 30 runs, 11 had the green beard allele take over completely, and

在 30 次模擬中，有 11 次綠鬍子等位基因完全佔了上風，

Even though the greenbeard allele had an advantage and saw its frequency rise on average, it's still subject to the whims of luck when there are only a few copies.

即使綠鬍子等位基因有優勢，平均來說頻率也會上升，但當數量很少時，它仍然會受到運氣的影響。

And once extinction happens, it's permanent.

It's interesting to imagine what kinds of alleles might have appeared in the history of the Earth and could have done well, but

想像一下，地球歷史上可能出現過哪些等位基因，並且表現得很好，但

The second interesting thing is that it seems like the green beard allele's advantage depends on how many greenbeards there are.

第二件有趣的事是，綠鬍子等位基因的優勢似乎取決於有多少綠鬍子。

Each run starts out with mostly random wobbling, but

每次運行都以大部分隨機波動開始，但

when the density of greenbeards gets higher, the helping happens more often and they get more benefit.

當綠鬍子的密度變高時，互助行為發生得更頻繁，它們也獲得了更多好處。

And in the 90% death chance case, things are just really chaotic.

而在 90% 的死亡機率情況下，情況非常混亂。

But we still do see an upward trend on average.

但我們平均來看仍然可以看到一個上升的趨勢。

字幕列表影片播放

模擬綠鬍子利他行為！ (Simulating Green Beard Altruism)

process

scenario

inevitable

eliminate