Two cents.

Almost all libertarians agree that free will requires these three things, which, as far as we can tell, hold true in this universe (mechanisms can differ):

1. Negation of complete determinism with more than one compatible outcome per given state, such that..
2. A form of causality or partial determinism holds which links cause and effect, and which makes possible..
3. Choice or collapse of outcomes, that can be meaningfully attributed as up to the agent

Negation of determinism means "not all events are determined", leaving room for events that are completely determined and also room for law-like processes. These law-like processes implement a form of causality which is necessary if choice is to be meaningful. In this sense, known physical laws can act as a suitable form of causality as long as they do not amount to complete determinism (ie standard interpretatiosn of basic physics are fully in accord with this).

Events not being predetermined are random, however randomness, in this sense, does not mean nor lead to out of one's control. Randomness in the sense of negation of complete determinism is not, in principle, incompatible with agent control and agent choice. Choice, so construed, can be meaningfully attributed to an agent in various ways both objective and subjective (a survey of ways is out of the scope of this answer).

Finally, the agent's reasons, preferences, tastes, goals and whatnot of course play a role in the choice made. Some people misunderstand this in the sense that they take this role to be always fixed, thus the choice is completely determined. The answer is that this role of agent's preferences is not fixed nor static (eg agent can re-prioritise her reasons, or even choose the same thing for different reasons, etc..), nevertheless it is present.

The thought experiment in @MichaelDorfman's answer is somewhat misleading. The reason is that it purports to tell apart free will from randomness and deterministic randomness via a black box experiment. We know that if free will exists it is or can be as random as any random process. Similarly it can be as orderly as any law-like process. So the proposed experiment is not a good way to argue about free will (or lack of it) because neither its observed randomness nor its observed law-likeness is what tells it apart. Rather, as argued here, it is the plausible conditions that make (libertarian) free will possible, plus ways to meaningfully attribute it to the agent in principle.

Two cents.

The thought experiment in @MichaelDorfman's answer is somewhat misleading. The reason is that it purports to tell apart free will from randomness and deterministic randomness via a black box experiment. We know that if (libertarian) free will exists it can be as random as any random process (eg simply choose to copy any random process). Similarly it can be as orderly as any law-like process (eg simply choose to copy any algorithmic process). So the proposed experiment is not a good way to argue about free will (or lack of it) because neither its observed randomness nor its observed law-likeness is what tells it apart. Rather, as argued here, it is the plausible conditions that make (libertarian) free will possible in the universe, plus ways to meaningfully attribute it to the agent in principle.

Almost all libertarians agree that free will needs three things, which, as far as we can tell, hold true in this universe (exact mechanisms can differ):

1. Negation of complete determinism with more than one compatible outcome per given state, such that..
2. A form of causality or partial determinism holds which links cause and effect, and which makes possible..
3. Choice or collapse of outcomes, that can be meaningfully attributed as up to the agent

Negation of determinism means "not all events are determined", leaving room for events that are completely determined and also room for law-like processes. These law-like processes implement a form of causality which is necessary if choice is to be meaningful. In this sense, known physical laws can act as a suitable form of causality as long as they do not amount to complete determinism (ie standard interpretations of basic physics are fully in accord with this).

Events not being predetermined are random, however randomness, in this sense, does not mean nor lead to out of one's control. Randomness in the sense of negation of complete determinism is not, in principle, incompatible with agent control and agent choice. Choice, so construed, can be meaningfully attributed to an agent in various ways both objective and subjective (a survey of ways is out of the scope of this answer).

Finally, the agent's reasons, preferences, tastes, goals and whatnot of course play a role in the choice made. Some people misunderstand this in the sense that they take this role to be always fixed, thus the choice is completely determined. The answer is that this role of agent's preferences is not fixed nor static (eg agent can re-prioritise her reasons, or even choose the same thing for different reasons, etc..), nevertheless it is present.

Two cents.

Almost all libertarians agree that free will requires these three things, which, as far as we can tell, hold true in this universe (mechanisms can differ):

1. Negation of complete determinism with more than one compatible outcome per given state, such that..
2. A form of causality or partial determinism holds which links cause and effect, and which makes possible..
3. Choice or collapse of outcomes, that can be meaningfully attributed as up to the agent

Negation of determinism means "not all events are determined", leaving room for events that are completely determined and also room for law-like processes. These law-like processes implement a form of causality which is necessary if choice is to be meaningful. In this sense, known physical laws can act as a suitable form of causality as long as they do not amount to complete determinism (ie standard interpretatiosn of basic physics are fully in accord with this).

Events not being predetermined are random, however randomness, in this sense, does not mean nor lead to out of one's control. Randomness in the sense of negation of complete determinism is not, in principle, incompatible with agent control and agent choice. Choice, so construed, can be meaningfully attributed to an agent in various ways both objective and subjective (a survey of ways is out of the scope of this answer).

Finally, the agent's reasons, preferences, tastes, goals and whatnot of course play a role in the choice made. Some people misunderstand this in the sense that they take this role to be always fixed, thus the choice is completely determined. The answer is that this role of agent's preferences is not fixed nor static (eg agent can re-prioritise her reasons, or even choose the same thing for different reasons, etc..), nevertheless it is present.

Two cents.

Almost all libertarians agree that free will requires these three things, which, as far as we can tell, hold true in this universe (mechanisms can differ):

1. Negation of complete determinism with more than one compatible outcome per given state, such that..
2. A form of causality or partial determinism holds which links cause and effect, and which makes possible..
3. Choice or collapse of outcomes, that can be meaningfully attributed as up to the agent

Negation of determinism means "not all events are determined", leaving room for events that are completely determined and also room for law-like processes. These law-like processes implement a form of causality which is necessary if choice is to be meaningful. In this sense, known physical laws can act as a suitable form of causality as long as they do not amount to complete determinism (ie standard interpretatiosn of basic physics are fully in accord with this).

Events not being predetermined are random, however randomness, in this sense, does not mean nor lead to out of one's control. Randomness in the sense of negation of complete determinism is not, in principle, incompatible with agent control and agent choice. Choice, so construed, can be meaningfully attributed to an agent in various ways both objective and subjective (a survey of ways is out of the scope of this answer).

Finally, the agent's reasons, preferences, tastes, goals and whatnot of course play a role in the choice made. Some people misunderstand this in the sense that they take this role to be always fixed, thus the choice is completely determined. The answer is that this role of agent's preferences is not fixed nor static (eg agent can re-prioritise her reasons, or even choose the same thing for different reasons, etc..), nevertheless it is present.

The thought experiment in @MichaelDorfman's answer is somewhat misleading. The reason is that it purports to tell apart free will from randomness and deterministic randomness via a black box experiment. We know that if free will exists it is or can be as random as any random process. Similarly it can be as orderly as any law-like process. So the proposed experiment is not a good way to argue about free will (or lack of it) because neither its observed randomness nor its observed law-likeness is what tells it apart. Rather, as argued here, it is the plausible conditions that make (libertarian) free will possible, plus ways to meaningfully attribute it to the agent in principle.