Can the goals of an organism be imputed from observation?

Question

For example: we observe an ant carrying food back to its nest. We may speak like the ant has a goal of increasing the amount of food in the nest. We observe a student proofreading an essay. We may speak like the human has a goal of improving the grammar and wording of the essay.

The most explicit definition of "goal" I can give is this: a goal is an outcome that an entity is seeking. If the entity chooses actions that the entity believes will increase the chances or the value of the sought outcome, then the sought outcome is a goal.

To determine what goal the entity is seeking, we may observe first what the organism is doing, and second, how the organism changes its behavior in response to some disruption. If we place an impassable barrier in front of the ant, does the ant change its behavior to return to the nest by a different path? If we introduce spelling errors into the essay the student is proofreading, does the student fix them? Generally, if we suspect a system is seeking a goal, the way to test this suspicion is to put some obstacle in the way of the goal and see if the system can compensate for it and head towards the goal anyway.

The above concept is that of an attractor. If we perturb the system a little bit away from a region, does the system return to the region?

There is an objection that many systems tend towards attractors without being intelligent decision-makers. A pendulum with friction tends towards a stationary vertical position as an attractor. Let's just set this aside for now, and suppose that we are only trying to determine the goals of a system with some minimal intelligence, and not applying the methods to simple systems like pendulums.

There is an objection that the process of testing might take an excessively long time period, especially if the organism has secret long-term goals that they are deliberately hiding. But we are asking about what can be ideally found through observation, rather than with the limits of our current technology. With advanced enough technology we could measure the organism to such precision and detail that we can build a neural model of its brain, so that with a powerful enough computer we can answer all subjunctive questions about what it would do in different scenarios, without having to wait for long time periods. This is still empirical observation because the model was derived from empirical observation, and we'd still be looking for the attractors in the organism's behavior.

Let us consider specifically that we encountered an intelligent alien organism, so we have no prior reference for their behavior and cannot rely on metaphors with humans. The only thing we can do is look at its behavioral attractors.

So, is it generally possible to impute the goals or intentions of an intelligent creature from its behavior and its (subjunctive) reaction to setbacks, thus reducing "intention" or "will" to a matter of empirical observation?

Answer 1

Up to a point, yes, but the visual aspects of a person's behaviour may contain a degree of ambiguity. If you see me reaching for a medical textbook, am I studying for an exam, or researching to write a novel, or trying to diagnose symptoms I posses, or trying to diagnose symptoms someone else possesses, or planning to poison my annoying neighbour, or looking up an answer to a crossword clue, or compiling a pub quiz, or picking up an object to prop-up a wonky tripod, or etc etc. I am sure you can think of some other possible goals I might have in mind when reaching for a medical text. You might respond to my challenge by saying that were you to observe the behaviour for a longer period, much of the ambiguity might fall away. True. But I can also play the trick of extending the temporal period under consideration. Suppose I am study for an examination. Is my 'goal' to pass the exam, or is it to keep my parents happy, or is it to earn a ton of money from offering plastic surgery to the image obsessed, or is it to infiltrate the army medical corp as a Russian spy, or is it... etc etc.

In summary, yes you can probably figure out what someone is trying to achieve from looking at what they are doing. Who would have thought it?

Answer 2

No. Because the nature of minds is the purposes can change and update through internal processes which cannot be observed from behaviour.

Consider how evolution doesn't just 'know' what to do, it tries things, and successful things remain as impulses to behaviour. If you can repeatedly observe a species in the correct circumstance to trigger that, you will be able to see the behaviour helping the animal survive and reproduce. But if it doesn't get triggered, you can't, and you won't know what's there to get triggered when. A good example might be cats that get terrified by cucumbers (links to an article by a veterinary behaviourist, about the topic). Only through observation of many cats with many objects could you stumble on that. And you can't know what hidden behaviours there are for what triggers.

More generally, research on post hoc reasoning shows humans often jump without reasoning to conclusions unaware consciously of why, and only then use reasoning to find justifications. We see this in the Is-Ought distinction in moral reasoning, in encounters with 'moral dumbfounding' (links to journal article about). We can look for explanation to Kahneman's Type 1&2 Reasoning, that full scrutiny is resource intensive and slow, and much of life requires getting on with 'good enough' responses until they fail or we face unknowns.

You can also look at the total failure of the Behaviourism research paradigm in psychology (link is to Wikipedia section). I would point to ideas about intersubjectivity, and cognitive models for social prediction evidenced by the Dunbar Number and the Default Mode Network, to understand how we go from observing the behaviour of others on to 'getting inside their heads'.

You might also look towards the idea of Strange Loops, where a systems ability to carry a model of itself in it's reasoning, to shape it's choice how to be, introduces unpredictable feedback loops that are no longer subject to simple forms of logic (ie First Order logics).

You are interested in Strange Attractors, so you might like to look into leading complex-systems theorist David Krakauer's framing of 'teleonomic matter', eg as discussed in this Mindscape podcast episode: Complexity, Agency & Information. It's an approach which can derive 'purposes' of living systems and non-living ones from the same framework.

Answer 3

There are two different questions being asked here. The one in the title, can we impute goals to organisms based on behavior, is a clear yes. Behavior provides the main data we work off of for imputing goals.

The second question is at the end of the discussion, can goals be REDUCED to behavior, as per the discussion of attractors, inhibitors, etc. The answer on this reductionism is a NO. And this is because, to infer goals, we are NOT only looking at behavior, we are also looking internally, at our own minds, and extrapolating off our internal intentionality and inferring a theory of mind for other agents too. This internally based inference is NOT just a replication of our own minds, but also involves speculative thinking about how other minds can be different from our own, is a significant secondary source of data, and is what allows us to understand complex motives, and deception.

The attempt to reduce mind to behavior alone is what was central to the philosophic and psychological movement called behaviorism. As a psychological theory, behaviorism is really only useful for very simple creatures that do not have awareness of the intentionality of other creatures. Behaviorism tried to do psychology without Theory of Mind, and has mostly been abandoned as an extended bizarre failure. Since awareness of the intentionality of other creatures is central to psychology it was basically an effort to abandon/ignore the psycho part of psychology.

Answer 4

Two cents.

We know goals (eg as mental states) exist because we have goals and experience ourselves acting towards satisfying those goals (although this does not mean we always succeed).

From this observation, about our goals, we can extrapolate, and assume other living systems can have goals as well (for example the closest species to us, some apes).

From this we can extrapolate more, only with some risk, and assume many or all living things have goals (although the exact experience of those goals may vary widely). This type of extrapolation is open enough to allow for variation that is quite plausible.

The OP tries to represent goals as attractors (of behavior). This is plausible, and by known results on dynamical systems (ie KAM theory) we know that a system perturbed tends to stay close to the original behavior given the perturbation is within some limits.

Can we attach goals to an alien creature based on observing their behavior?

My opinion is that it is possible under two conditions: a) a sufficiently long and varied period of observation which, b) implies without reasonable doubt that the best explanation for what observed is the presence of certain goals.

Answer 5

Not in general, because not everything has goals. Many processes have effects that we can recognize, but that doesn't mean they were explicit goals.

For example, sunshine and rain (indirectly) cause plants to grow. But we would not say that this was their goal. They are spontaneous processes, and plants simply take advantage of them.

Ancient civilizations did not have the same perspective, though. They inferred intentionality everywhere, and formulated the existence of gods that controlled these processes; the gods had goals and could be influenced through prayer and sacrifices.

Now we understand physical processes somewhat better. Many of them (like rain) just happen. We generally only impute goals to living matter, but even there it can be difficult to distinguish goals from unintentional side effects. When a bee flies from flower to flower, it gathers nectar and pollinates the plants; we usually consider the first to be a goal (because it directly helps the bee) and the second to be a side effect, but this is somewhat arbitrary. Bees and flowers co-evolved to take advantage of each other -- bees benefit from pollinating plants so there will be more of them to get nectar from.