Hacker News Reader

Comment by jacinda | Hacker News Reader

Actually, they have a control in place in the study itself and they could additionally have compared against the results from the 1988 study where all the experimenters hands were free.

Sometimes the experimenter had their hands free and other times they did not. When the experimenter had their hands free and tapped the box with their forehead, the babies imitated the head action over 60% of the time. When the experimenter had their hands bound, the babies imitated the head action only 20% of the time.

To me that's a big enough difference that the babies didn't just accidentally tap without specific intent.

Actual write-up in Nature (which was linked to in the article): http://www.nature.com/articles/415755a.epdf

View on HN

>> Actual write-up in Nature (which was linked to in the article):

Yeah, about that:

79% of them chose not to imitate here because their own hands were free, presumably concluding that the head action was the most rational.

Whether they re-enacted the head action or not, all infants who watched the adult perform under both conditions still used the hand action.

See what I mean? Who knows _why_ 79% of babies chose to imitate the researcher? The research assumes it was "because their own hands were free" - but isn't that what it's trying to prove in the first place? And where does that conclusion, about rationality, comes from? Certainly not something that the babies indicated, but rather the researchers' interpretation of the actions of the babies. This is... it's kind of like anthropomorphising animals. There's absolutely no reson to assume that babies where behaving rationally, or irrationally. They could just be moving completely at random, but the researchers go ahead and assign motives to them and "rationality" as well. What the hell!

And like the second paragraph says, all infants used their hands _anyway_. At that point it's obvious that the head action is just a random thing that some babies do anyway. Because if they know they can turn the light on with their hands, why would they use their heads _also_? If they don't know they can turn the light on with their hands, then how can they know they can turn it on with their heads? This just doesn't make any sense.

It feels like you are shifting the goalposts here, but I potentially misunderstand your original objections.

In your first comment, you say that "maybe the babies had just completely forgot about the experimenter and his wrapped-up hands and didn't even "tap" the box but more like "bumped into" it hands-first". That is, we can't judge the intent of the babies and so we can't determine why they use their hands instead of heads.

I agree that we can't ask babies what their intent is and have them verbalise the response, but that is not the same thing as being unable to determine intent. Indeed, asking and listening to a response is just one form of setting up an experiment and evaluating the response. Of course there may be reasons we can't think of, but the scientific method gives us a framework to explore those ideas.

Hypothesis: babies can recognise constraints upon action when observing action. When mimicking this action later, if not under the same constraint, the babies will modify the action in order to achieve the outcome of the action as simply as possible.

The null hypothesis here would be every baby performing the same action regardless of constraints observed.

So, if after seeing the button pusher with hands tied, and at other times without hands tied (using different babies in each case of course), we would expect to see no difference in how the babies pushed the button. If they were simply "bumping into" the button with their hands, they would do so an equivalent number of times in each group of babies. Assuming everything else was controlled for (which we can't know for certain, but is reasonable to assume) then the hypothesis holds.

If the babies were "moving completely at random" we would expect to see completely different outcomes then what were observed.

Your objection about rationality is probably more about scope or context of definition. The researchers seem to be using 'rational' as a shorthand for 'understands constraints and can remove them in order to achieve the desired outcome'.

For example, if the observed uses their head to push the button and their hands are not tied, it is 'rational' to believe that the head must be used to push the button. If the hands are tied, however, it is more 'rational' to believe that the head is being used because the hands are tied - the constraint is recognised.

Finally the second paragraph says "all infants who watched the adult perform under both conditions" not "all infants". That is, there was a group of babies shown both ways to push the button, and those babies chose to use their hands every time and not their heads.

>> It feels like you are shifting the goalposts here, but I potentially misunderstand your original objections.

I have several objectsions about this and similar studies, so I may sound like I'm changing the subject, sorry about that. I'm not sure where I did that, but it's not my intention to shift the goalposts, I feel I'm being consistent.

>> The null hypothesis here would be every baby performing the same action regardless of constraints observed.

That's just not a valid null hypothesis. Not when you're dealing with living things, particularly human beings, even if they're babies. There's absolutely nothing to say what a person, even a baby, will do in a certain situation. We are not automata, or if we are some sort of automaton we are a type of automaton that is too complex for us to understand and predict.

In that sense I don't see how there can even be a valid null hypothesis. We don't even know what to expect in the first place, so what are we comparing some specific observation to?

The idea is that the study has "controls", but note we're not even given the numbers of the subjects the experiment involved. Even the bar chart on the right has percentages rather than numbers- why? Well, because it probably involved a dozen subjects or so. A couple of dozen? Probably not more than that.

We're talking about tiny, tiny numbers here. There's not really enough to seriously talk about "controls". And my intuition is that if the study was done on, say, a few hundred babies, results supporting the researchers' conclusions would dissolve into thin air. But- who knows?

As to the main hypothesis:

>> Hypothesis: babies can recognise constraints upon action when observing action.

Why is that a valid hypothesis to state in the first place? It's making some assumptions about the baby's mind, specifically that it's the same kind of mind, in the same kind of context, as that of the person who stated the hypothesis. Well- that sort of assumption is something that has to be tested in the first place. We can't just hand-wave that fundamental question away with a "duh, it's a human baby so".

Social sciences have a tendency to make gigantic leaps of reasoning like that, and bridge the gap between observed behaviours with assumptions and common sense, or other social scientists' theories. There are so many things that is probably impossible to evern know for sure about our minds, no matter how much science we throw at them, or at the very least it will takes us generations upon generations of painstaking research to figure out- yet, a lot of the people who study minds and behaviour will gladly take them for granted and publish a paper taking them for granted, then others will cite that paper, build upon its baseless assumptions and add their own on top. That's not science- that's turtles all the way down.

>> there was a group of babies shown both ways to push the button, and those babies chose to use their hands every time and not their heads.

I do think that's what I meant but even so- what does that tell us? Who says that's not what should be expected? We're back to forming a valid null hypothesis, and I don't see how that is established.

> There's absolutely nothing to say what a person, even a baby, will do in a certain situation.

Perhaps not for a single individual in a single situation, but on average humans are quite predictable, and not the unique little snowflakes you make them out to be. If someone slaps you in the face, odds are you will be angry, frightened, surprised, or some combination thereof. You're probably not going to hug that person and invite them over for dinner. Of course there are 6 billion people on this earth, so there may be someone out there who will start singing when slapped, but statistically that person doesn't exist.

> We are not automata, or if we are some sort of automaton we are a type of automaton that is too complex for us to understand and predict.

Understanding the circumstances, thoughts, and emotional state of others is required for empathy. If humans were truly too complex for other humans to understand and predict, empathy could not exist. And if people were unable to predict others, it would be impossible to troll people on the internet.

> In that sense I don't see how there can even be a valid null hypothesis. We don't even know what to expect in the first place, so what are we comparing some specific observation to?

By that logic there can never by a null hypothesis for any experiment. The world is a big and complex place, and there is pretty strong evidence that randomness plays a big role at the smallest scales that we currently understand. Which direction will this rock go when I drop it? Will it even move at all? Who knows! Through the power of quantum teleportation it could suddenly appear anywhere.

> The idea is that the study has "controls", but note we're not even given the numbers of the subjects the experiment involved.

Why are you worrying about the number of subjects? If human behavior cannot be predicted at all, we wouldn't really expect to see a pattern no matter how many subjects participate in the study, would we?

> Even the bar chart on the right has percentages rather than numbers- why? Well, because it probably involved a dozen subjects or so. A couple of dozen? Probably not more than that.

If you have a couple of dozen subjects, observing something happening less than 20% of the time in one group and more than 60% of the time in the other is actually pretty significant. You don't need hundreds of subjects when the difference is that big. [This page](http://www.nature.com/nature/journal/v415/n6873/fig_tab/4157...) seems to suggest that one group had 14 participants, and the other 13. I suppose that technically qualifies as "couple of dozen".

> We're talking about tiny, tiny numbers here. There's not really enough to seriously talk about "controls". And my intuition is that if the study was done on, say, a few hundred babies, results supporting the researchers' conclusions would dissolve into thin air. But- who knows?

I'm sure researchers everywhere would absolutely love to have hundreds of people participate in all their studies. But that's expensive. Let's imagine a world where this study was done on 200 babies. People in that world would still complain about this study! Instead of "27 subjects? Surely that is too few to conclude anything!" we would hear "How dare those researchers waste a million dollars to investigate how often babies smash stuff with their head!". Imagine the study was unable to show any result, the outrage would be terrible.

> Why is that a valid hypothesis to state in the first place? It's making some assumptions about the baby's mind, specifically that it's the same kind of mind, in the same kind of context, as that of the person who stated the hypothesis. Well- that sort of assumption is something that has to be tested in the first place.

I fail to see your objection here. The hypothesis would make sense if the subjects were adults. It would also make sense for children. So at what age does it stop making sense? How young does someone have to be before we can no longer make assumptions about their mind? How would we even know without experiments such as the very one you are now criticizing? As someone who used to own a dog I think the hypothesis would even make sense not just for humans but also for some social animals.

> We can't just hand-wave that fundamental question away with a "duh, it's a human baby so".

Well if I had to choose between "the mind of a human baby functions similarly to the mind of a human adult" and "the mind of a human baby is completely alien and nothing like that of a human adult" I think Occam's razor favors the former. Regardless, your approach leads to some type of analysis paralysis: we can't perform a study without a hypothesis, and we don't have enough data to form a hypothesis.

> Social sciences have a tendency to make gigantic leaps of reasoning like that, and bridge the gap between observed behaviours with assumptions and common sense, or other social scientists' theories.

So in that respect it's a lot like the non-social sciences?

> There are so many things that is probably impossible to evern know for sure about our minds, no matter how much science we throw at them, or at the very least it will takes us generations upon generations of painstaking research to figure out-

That seems like a giant leap of reasoning of the same type that you like to criticize the social scientists for. Do you at least have a study that vaguely supports this theory?

> yet, a lot of the people who study minds and behaviour will gladly take them for granted and publish a paper taking them for granted, then others will cite that paper, build upon its baseless assumptions and add their own on top.

That sounds a lot like science works in practice.

> That's not science- that's turtles all the way down.

Well there's only so far you can build on a false theory before you notice that your data doesn't make any sense.

> I do think that's what I meant but even so- what does that tell us? Who says that's not what should be expected? We're back to forming a valid null hypothesis, and I don't see how that is established.

I suspect that is primarily because you're being deliberately obtuse because of your dislike for the social sciences.

>> When the experimenter had their hands free and tapped the box with their forehead, the babies imitated the head action over 60% of the time.

There's an episode in Friends where Phoebe discovers she can change channels on the TV by blinking (if I recall correctly). Every time she blinks, the TV switches channel. Joey is there and he confirms it: whenever Phoebe blinks the TV changes channel. Every single time. Amazing. Then she loses it.

So, I'm not saying you have to be Phoebe to get in that situation, in fact you certainly don't. It's very hard to untangle the mess of causality when you're experimenting, especially if we're talking humans, even infants, who are incredibly, impossibly complex.

In any case all this is completely besides the point: it's impossible to know another person's mind state with accuracy unless you ask them and even then it's really hard to know for sure. So any experiment that purports to prove something about a person's mind state is most probably making a mistake, somewhere- same as papers purporting to show energy from the void and so on.

Reply on news.ycombinator.com