Can libertarian free agency be distinct from deterministic, random, and semi-random causation?

Today’s debate prompt is summarized in the title and fully stated at the end of this post.



I ask you to presuppose that the physical world is indeterministic, that there is a mental substance (minds are nonphysical), and that there are at least three types of causation: deterministic event causation, whereby the outcome of some event may be predicted with absolute certainty according to physical laws; random event causation, whereby physical laws allow for multiple valid yet mutually exclusive outcomes to some event, and where even with perfect physical knowledge the outcome may be most accurately predicted by assigning equal probabilities to each physically possible outcome; and finally semi-random event causation, which is a combination of the two such that given perfect physical knowledge, the most accurate predictions rely on disuniform probability distributions.

So to define terms, an event is a process that produces an outcome. Outcomes, groups of outcomes, and groups of events are events in their own right. Causation refers to an event producing an outcome. Events (and by extension, outcomes) may be classified according to causation. Every outcome is caused by an event, and every outcome in turn is part of at least one event causing its own outcomes, to make a “causal chain” of events causing outcomes which are events causing outcomes which are events causing outcomes, &etc.

I ask you to presume for this debate that not all events are the outcomes of other events. An event which is not the outcome of any other events is an “uncaused” event. For example, all random events are necessarily uncaused. There is a special case of uncaused deterministic event causation (only one outcome, no causes). To clarify, events can be uncaused, but all outcomes have causes.

As a matter of style I will be using capital letters for events (“event A”, “event B”) and italicized capital letters with a number to differentiate outcomes to a particular event (“outcome A1”, “outcome A2”).

Examples of event causation

Let’s put this into practice before going into the debate proper. Suppose event A randomly leads to either outcome A1 or A2, while event B randomly causes either outcome B1 or B2 with equal probability. Now suppose event C is a deterministic process acting on the outcomes to random events A and B. Physical laws allow us to definitively predict the outcome of event C if we are given the outcomes to A and B: if we start with A1 and B1, we get C1; A2 and B1 also gives C1; A1 and B2 gives C2; and finally A2 and B2 gives C3. Refer to the following tables:

Event A, random causation
A1 (50%) A1 (50%)

Event B, random causation
B1 (50%) B1 (50%)

Event C, causally determined by events A and B
A1 A2
B1 C1 C1 => C1, C2, or C3
B2 C2 C3

Event C is deterministic, and events A and B are both random. Now suppose I grouped all three events together, and labeled the whole cluster “event G”. Treating a whole group of events as a single event is a metaphysical property events, or at least I ask you to assume it is. Outcomes C1, C2, and C3 would correspond to outcomes G1, G2, and G3 respectively. Would event G be deterministic, random, or what?

Event G - semi-random causation
A1 (50%) A2 (50%)
B1 (50%) C1 (25%) C1 (25%) => G1 (50%) G2 (25%) G3 (25%)
B2 (50%) C2 (25%) C3 (25%)

According to probability theory, event G is 50% likely to lead to outcome G1, 25% likely to lead to outcome G2, and 25% likely to lead to outcome G3. Because the outcome to event G cannot be definitely predicted, event G is not a deterministic event. Because the outcomes are not equiprobable, event G cannot properly be described as a random event (as defined above, random events have equiprobable outcomes). Event G belongs to that third category of events, which I call semi-random. I define semi-random events as random and deterministic events grouped together, leading to multiple possible outcomes where - even with perfect physical knowledge - the outcome can be most accurately predicted only by assigning disuniform probabilities.

What is an agent?

An agent is a person, or at least the mental state or soul that makes him or her a person. I ask you to assume an agent may not be reduced into physical pieces or processes.

What is free agency?

Free agency in the libertarian philosophy refers to an agent’s ability to make and carry out a decision without being forced into a choice by external influences. In study of law ‘being forced into a choice’ might include coercion, but for metaphysical purposes the meaning is more narrow: literally removing the choice from the agent’s power. Libertarian free agency is the ability for an agent to do otherwise in exactly the same situation.


I’ve attempted to define deterministic causation, random causation, semi-random causation, and libertarian free agency. For debate, can libertarian free agency be a distinct form of event causation?


One argument, in favor

If I carry out some decision and later feel regret/pride, the intuitive assumption is that I actually had the ability to do otherwise but did not do so. At that moment in the past, I had the ability to take another action. Nothing forced me into a choice, not the laws of physics, not random chance, not even some long string of actions that lead back to before I was born.

I know my feelings exist, but feelings are not hard evidence as to the existence of free agency. Free agency could very well be an illusion, and I acknowledge a leap of faith is necessary to conclude that it is real. On the other hand, all inductive logic requires a leap; all higher knowledge is built on inductive logic. I defy anyone to prove (perhaps in a dedicated topic) the existence of randomness as a metaphysical concept without making a leap comparable to the one I must make to believe in free agency.

Given my preexisting beliefs, I’ll attempt to argue that libertarian free agency can exist as a distinct form of event causation.

Suppose a person must make a decision from multiple mutually exclusive choices. The decision is an event and each choice is an outcome. I might refer to the outcomes of decisions made by agents as ‘actions’. If the person has free agency, he or she must be able to make and carry out the decision without being forced into a choice by external influences.

Let the decision exemplifying free agency be event D (for “decision”), with possible outcomes D1, D2, D3, and D4. Remember that the agent making the decision is a person, or at least their mental state which makes them a person. As such he or she, in that state, is the outcome of a variety of previous events. Let that set of events be notated as event A (for “agent”), and therefore let the particular agent we are interested in be notated as A1. Said agent likely relies on environmental factors to make the decision. The environment is itself the outcome of a set of events which we will label event E (for “environment”). Let the state of the environment we’re interested in be notated E1.

Event D - free agent causation,
given events A and E
E1 E2 ...
A1 D1, D2, D3, or D4 ... ...
A2 ... ... ...
... ... ... ...

The free agent has, at the very least, the physical ability to choose and carry into effect either action. If it were not so, the laws of physics would force the agent into making a single, predictable choice. Deterministic event causation precludes all but one outcome, therefore this definition of free agency is incompatible with determinism. Event D cannot be deterministic.

Could event D be random? This will depend on the physics. I suspect the answer is ‘no’, because I don’t think any person’s choices are equiprobable in all cases. At the macro level, people are certainly not random. But it is possible that the only nonphysical interactions are totally random, and that the physics “smooths out the randomness”; that the only indeterminism in the world is such that all human action at a macro level is semi-random. On the other hand, such a scenario sort of ignores the substance dualism I wrote into the debate. If the concept of a nonphysical mind (or soul) is of any significance whatsoever, they would not make all free decisions with such indifference that we could accurately assign equal probability to all physically possible outcomes.

It is no difficult task to imagine event D as semi-random. In the same way that semi-random events are assumed to exist in physical processes, we could generalize the same for mental processes. If free agency is semi-random, we run into the same problem - is my ‘self’ nothing but random chance being smoothed out by physics? There is no way to falsify this generalization, but at the same time, it is not necessarily justified. A property of the mental substance is that it is a sort of black box. No physical experiment can reveal the inner workings of a mind. No amount of physical knowledge can guarantee that an observed non-uniform probability distribution is the most accurate way to predict an agent’s decisions.

And so, finally, could event D be some other category of event causation? I think it could be, if instead of assuming the outcome is bound by deterministic or random processes, I admit that even with perfect knowledge of the physical world, A1 is a black box, the actual mental process is unknowable, and the actual probability of each outcome is undefined. This is consistent with my idea of libertarian free agency: the agent is not forced to make any particular decision, not even over a long run.

Event D - free agent causation,
given events A and E
E1 E2 ...
A1 D1, D2, D3, or D4;
probabilities undefined
... ...
A2 ... ... ...
... ... ... ...

You can draw up probability distributions based on previous and similar trials, but these are never guaranteed to be accurate; as a matter of philosophy, the agent has the power to defy your expectations. Formally speaking, if A1 and E1 converge for event D the probability of outcome D1 is undefined, as are the probabilities of outcomes D2, D3, and D4.


Wow, such long posts. I think if you spent even a fraction of this time thinking about how decisions are made under your libertarian free will concept, you’ll appreciate the critical problem with it. A knowing, intentional decision is always going to be based on the past and current state of the universe. Regardless of what we include in “universe” the problem remains the same.

Some more specific points:

I regret playing Bishop to e4 even though it seemed the best move based on my observations and understanding at the time.
But rewind the tape, including rewinding my brain, soul, whatever to it’s prior state, and I’ll play Be4 again.

(Of course I would also say I was “able” to play other moves because I use the word “able” in the normal way, which implies nothing about indeterminacy)

I think “prediction” should be put to one side. It’s the thing that gives people a hangup about Determinism, but as I’ve pointed out, there are logical problems with an agent in the universe being able to make predictions about the universe, including paradoxes.

So it’s better to describe Determinism in terms of the future state being determined from the current state, rather than in terms of prediction when the latter is not only not an implication but could be said to be explicitly physically impossible.

Does your soul have certain properties? For example, might one man’s soul be more predisposed to violence than another’s? If not, then you’ve shot your own theory in the foot, because environmental or biological factors are deterministically affecting the decisions your soul makes, which eliminates the idea of free will.

And if different souls ARE predisposed to different sorts of actions, you just kicked the determinism can down the road. Joey isn’t being an asshole of his own free will, he’s being an asshole because he has a shitty soul that’s making him act like an asshole.

The OP defines a ‘mental substance’ that is a detached observer of the physical world. Such a mental substance has free agency by definition.

So, libertarian free agency can be a distinct form of event causation if your premise of an independent mental substance is valid.

Hence why A1 (the agent) and E1 (the environment) converge for event D (the decision).

I’m not sure if you’re working within my framework (what I wrote above) or just expressing your general opinion on the subject of free agency, but I’m confident that the conclusion does not follow within my framework.

I prefer not to rely on the word “determined” when defining “determined” or “deterministic”. Only agents (real or hypothetical) make determinations, and whenever you say a “state” is determined by another “state”, you are personifying the state. I may as well be straightforward and say, given perfect physical knowledge, the outcome can be predicted according to physical laws. Or even better, just that the outcome could be predicted according to physical laws, the implication being that physical laws require a relevant degree of physical knowledge. Whether a person could acquire the necessary physical knowledge to actually make the prediction is irrelevant to the accuracy of my definition; we are discussing metaphysics, after all.


If it does, they are not defined by me in this topic and I don’t use them in my argument for a distinct classification of event causation.

I allowed for environmental and biological factors to affect the decisions an agent makes by including the environment as an input for the decision event.

I did not, however, admit that the decision event was deterministic and in fact concluded (as you do) that libertarian free agency is incompatible with a deterministic decision event.


The post is not correct.

At no point could perfect physical knowledge predict an event. Perfect physical knowledge after the fact would only allow the observer to understand the necessary conditions that eventually resulted in the observed event.

There is a semantic problem with the term ‘cause’. The definition of cause is temporal:

Prior to the event - At any point prior to the event the data set is not complete, As long as Event E1 is unknown - the set of physical conditions are not directed to a goal - they do not reveal a result (event). To be complete the data set must include the event. No matter how accurate, the data set alone is never predictive.

After the event - Event E1 is known - It’s cause can be traced to physical conditions.

Perfect knowledge of conditions cannot predict an undefined event. Posterior Analytics, Aristotle

Having said that - the evening of hurricane Sally I assured the wife that we were safe because the storm was coming ashore west of Escambia Bay, It would have to halt it’s path, turn 90degrees and run due east to impact us. That’s what it did and we woke up with 2 feet of water in the house. However my speculation was not predictive. None of the computer models predicted that path because their data set was, as yet, incomplete. And the match to my speculation could not be made until after the event was known.

My point was that such a prediction cannot interact with the universe in any way because it leads to an infinite recursion of the calculation of the universe’s future state needing to be an input to the calculation.

Of course we could argue that from a gods eye view such a prediction is possible. But I would say it’s a very misleading framing of Determinism to say it’s based on being able to predict some future state. Not being able to do something in this universe is normally synonymous with that something being impossible.

If you think anything else could happen, then let’s hear it. Rewind the state of my brain, soul etc, and how do I know not to play Be4? And what does it have to do with regretting the move Be4?

I’m sorry, Crane, I don’t understand why you would think so. I don’t understand much of your post…

The next entry in a data set can be predicted according to a pattern, for example.


CAN it? I content that any given set of numbers can actually be part of an infinite number of patterns. For example, [1,2,3…] could be followed by 4. It could also be followed by 1 [1,2,3,1,2,3…] or by 5 [1,2,3,5,7,9,14,17,20…]

Mijin said it better

Babale also.

The data set is a segment of a pattern, but it is not predictive.

There are practical examples of deterministic predictions, for example in newtonian physics or classical thermodynamics. You don’t necessarily need intimate knowledge of the entire universe in order to predict a localized event. Only relevant knowledge is required. I’m not sure why you bring up recursion.


If you know the pattern and you have enough partial data, you can predict the next entry.

If you know the (deterministic) physical law and you have enough physical data, you can make a prediction.


That is a premise all gamblers hold to be valid .

This is only true because the predicted state is known and is part of the calculation - per Mijin and Aristotle.

You are anticipating a known event rather than predicting an unknown.

I don’t understand why it is relevant that gamblers agree with me. :frowning:


Are not the words “calculate” and “predict” synonymous in this context? The reference to Aristotle is lost on me, and quite frankly, I don’t see any similarity in what you have written and what Mijin wrote.


Assuming this particular decision is one made according to libertarian free agency,

  • Your chess move is a decision event with libertarian free agency
  • Libertarian free agency is the ability for the agent to do otherwise in exactly the same situation
  • Therefore, your chess move is a decision event with the ability for the agent (you) to do otherwise in exactly the same situation

If you were to somehow go through the same decision event twice, it is not implied that the outcome will be the same.


An assumption in the OP is that “there is a mental substance (minds are nonphysical)”.

If that mental substance is contained within the physical universe, then you are correct, but the assumption is that it is not - it is nonphysical.

If you assume that the situation is reversed, and the physical universe is contained within the ‘mental substance’ then that argument doesn’t hold.