Partners and Patients: A Revised Grammar of Social Power

1

This brief comment concerns Arash Abizadeh’s (2021a: 13) recently proposed four-place ‘grammar’ of (agential) social power: that the social power of an agent, V, with respect to outcome (type) O consists in V’s capacity to effect (outcomes of type) O ‘with the assistance of agents X, despite the resistance of agents Y’. ¹ In proposing this grammar, Abizadeh rejects, first, a tradition, owing to Max Weber, that says that all social power is power to overcome resistance (Barry, 1988: 341; Weber, 1978: 53). And he rejects, second, a tradition, also owing to Weber, that says that all social power is ‘power-over’ others (Dahl, 1957: 202–203; Weber, 1978: 942).

I propose, in place of Abizadeh’s grammar, that V’s social power with respect to O consists in V’s capacity to effect O with the assistance of X, thereby affecting patients Y. This is a different grammar of social power, in two ways. First, it adds, so to speak, a variable-place concerning affecting others. Second, it strikes out the variable-place concerning overcoming resistance. This grammar thus goes a step further than Abizadeh’s in rejecting the tradition that says that all social power is power to overcome resistance. In part because not all power to affect others is power-over others, at least as Abizadeh understands ‘power-over’, this grammar also joins Abizadeh’s in rejecting the tradition that says that all social power is power-over others. It is offered in the spirit of a friendly amendment.

2

Having proposed that V’s social power with respect to O consists in V’s capacity to effect O ‘with the assistance of agents X, despite the resistance of agents Y (where either X or Y may be null sets)’, Abizadeh continues:

This yields three types of social power: the general case in which X and Y are both populated, and two special cases. Unilateral power arises when X is empty: this is social power-despite Y’s resistance (but without anyone’s assistance). Omnilateral power arises when Y is empty: this is social power-with X’s assistance (but with no capacity to overcome anyone’s resistance) (Abizadeh, 2021a: 13).

Abizadeh’s example of unilateral power is dictator-rule voting:

where there is one voter whose vote always determines the outcome regardless of how other voters vote. Whenever the dictator votes yes, she unilaterally effects a yes-outcome despite the no-voters (but without the assistance of the other yes voters: their votes are impotent) (Abizadeh, 2021a: 13).

Abizadeh’s example of omnilateral power is unanimity voting, where each voter has the power to effect O by voting for O with the assistance of every other voter voting for O, but where no voter has the power to overcome any voter who resists by voting against O. It is unanimity voting that most definitively shows that not all social power is power to overcome resistance. Importantly, non-social power, for Abizadeh, is the case ‘where both X and Y’ are ‘empty’.

Putting this in my own terms, we seem to have:

First Definition:

My power to effect O is social power iff

(i) Someone can assist me in effecting O – ‘X is non-empty’

or

(ii) I can effect O even if someone resists – ‘Y is non-empty’.

3

This First Definition implies that, implausibly, all power is social power: that there is no such thing as non-social power. Here’s an informal and rough argument.

4

We can put the argument more formally, laying out by the way a framework that will be used in the rest of the article:

Definition of circumstance

A circumstance is a possible situation that holds fixed at least everything apart from what I and other agents do.

Since they involve only one circumstance, stylized voting examples, on which we largely draw in what follows, are particularly convenient. Everything is fixed apart from how voters vote. We needn’t consider, as with, say, kicking a field goal, different circumstances involving different wind patterns, turf conditions and so on:

Definition of means

In circumstance C, I have a means m to O iff in C, m is an action I can perform such that, for some set of actions, n, that others can perform, if I m and others n, then I effect O partly because I m.

As I will understand things, and as Abizadeh (2021b: 306) agrees, I can effect O ‘partly because’ I m even if I my m-ing is not a but-for cause of O – even if it isn’t decisive.

Definition of power

I have the power to O iff there is a C such that in C I have a means m to O.

Definition of resisting someone’s effecting an outcome

In C, agent (or set of agents) Y resists my effecting O iff in C Y performs some action (or set of actions) r such that, for some means m* to O, if I m* and Y r’s and other agents perform some n*, then I do not effect O partly because Y r’s.

In other words, to resist someone’s effecting an outcome is to do something that might keep some means – something that could be effective, at least if others acted in the right way – from being effective. This fits the one passage where Abizadeh offers us guidance on how to understand resistance:

For example, in a voting structure with two possible outcomes, yes or no, in which a yes-outcome requires that an absolute majority of all eligible voters vote yes, abstaining from voting – forbearing – would unequivocally count as resisting the yes-outcome. This is because there are possible vote configurations in which the abstention would be decisive in securing a no-outcome. If the votes are just one shy of the quota for a yes-outcome, then the outcome would have been yes had a given abstainer voted yes (holding everyone else’s vote constant), that is, her abstention would be fully decisive for effecting the no-outcome (Abizadeh, 2021a: 12).

This example satisfies our definition of resisting someone’s effecting an outcome. Some Y (namely, the abstainer) performs an action r (namely, abstaining) such that for some means m* (namely, my voting yes) to O (namely, a yes outcome), if I take that means (vote yes), Y r’s (the abstainer abstains), and other agents perform some n* (namely, cast one more vote for no than for yes), I do not effect O (namely, the yes outcome is defeated) partly because Y r-ed (namely, because the abstainer abstained). Note that my voting yes counts as a means m* to yes outcome O since for some n, if others n (namely, cast just as many yes votes as no votes), I effect O partly because I m* (namely, I effect the yes outcome partly because I voted yes).

Putting this all together, we have:

First Definition, formal statement

My power to effect O is social power iff

(i) Someone can assist me in effecting O:

i.e., there is a C in which I have a m to O such that I effect O partly because I m only if and partly because for some agent (or set of agents), X, for some action (or set of actions), a, X a’s

or

(ii) I can effect O even if someone resists:

i.e., there is a C in which I have a m to O such that if I m, and for some agent (or set of agents) Y, Y resists my effecting O, and for some action (or set of actions) n, all other agents n, then I still effect O partly because I m. (The conditional here is to be understood as material.)

Now the formal version of the argument that, according to First Definition, all power is social power.

5

Presumably not all power is social power. Intuitively, my power to crack a coconut forthwith alone on a desert isle, when no one knows about me, no one would care about me if they did know, and no one can do anything with any chance of stopping me in time from cracking the coconut is not social power.

6

But this very example points the way to a response. The response is to reject the statement of (ii). Abizadeh might say, reasonably enough, that, when I effect O by m-ing in a circumstance in which there is no way for anyone to resist m, I don’t effect O despite someone’s resistance. There isn’t any resistance in the first place. Nor do I overcome anyone’s resistance. There is no resistance to overcome. So we should replace (ii) with (ii’) to arrive at:

Second Definition, formal statement

My power to effect O is social power iff

(i) Someone can assist me in effecting O:

i.e., there is a C in which I have a m to O such that I effect O partly because I m only if and partly because for some agent (or set of agents), X, for some action (or set of actions), a, X a’s

or

(ii’) Someone can resist my effecting O, and if they resist, I can overcome their resistance.

I.e., there is a C in which I have a m to O and some Y has a way of resisting, r, my effecting O, such that if I m, Y r’s, and other agents n, then I still effect O partly because I m.

My power to crack the coconut no longer counts as social power, because no one can resist my cracking the coconut.

7

The reader might wonder why we didn’t interpret Abizadeh as meaning (ii’) instead of (ii) in the first place. One reason is that (ii’) implies that Abizadeh’s illustration of unilateral social power, dictator-rule voting, is not a case of social power. Note that a dictator’s power is, like my power to crack the coconut, what we might call ‘unopposable’:

My power to effect O is unopposable iff for all C, for all m to O in C, in C, if I m, then I effect O partly because I m.

I.e., every means to O is sufficient in itself to effect O, no matter what others do.

The dictator has unopposable power because the dictator’s only means to effecting O is voting for O, and if the dictator votes for O, then the dictator effects O partly because the dictator votes for O (no matter what others do). In general, my unopposable power is not social power, by Second Definition. Since every means m to O is sufficient in itself, I can never be assisted. So (i) is not satisfied. Since every means m to O is sufficient in itself, I can never be resisted. So (ii’) is not satisfied.

There are two potential problems here. One is that we may wish, with Abizadeh, to count dictator-rule voting as social power. We will come back to this.

The other problem is that we might want a way to define unilateral power – intuitively power that can overcome any and all resistance without any assistance – such that it can be social. To address this second problem, let us say that:

My power to effect O is unilateral iff for all C, there is a means m to O such that if I m, I effect O partly because I m.

I.e., there is always a means that is sufficient in itself to effect O no matter what others do.

Whereas with unopposable power every means is by itself sufficient for O, with unilateral power every circumstance contains some means that is sufficient by itself for O. Unilateral power can nevertheless satisfy (ii’) and so be social according to Second Definition. This is because there can be other means that are not sufficient for O, such that I can be successfully resisted if I take those means. For an example of unilateral power that satisfies (ii’), consider:

Two Votes: Suppose that a majority is required for O. I have two votes and you, the only other voter, have one.

There is a way for you to resist. If I cast only one vote and cast it for O and you cast your vote against O, then I do not effect O. So (ii’) is satisfied. Nevertheless, there is a way for me to overcome your resistance: namely, by casting both votes for O. So my power is unilateral.

It will also be useful to define omnilateral power:

My power to effect O is omnilateral iff for all C, there exists X, a, such that for all m to O in C, I effect O partly because I m only if and partly because X a’s.

I.e., there is always assistance that I need to effect O, no matter what means I take.

This fits Abizadeh’s example of unanimity voting. There exists X (namely, any other voter), a (namely, voting for O), such that for all m to O (namely, my voting for O, which is my only means to O), I effect O partly because I m only if and partly because X a’s.

8

So, should we adopt Second Definition? The trouble is that there is unopposable power that is nonetheless intuitively social. Just imagine unopposable power to effect O where O affects others significantly. Suppose that on my desert isle I can forthwith cast spells that maim others, blight their crops, render them infertile and so on. Perhaps we should say that what makes my power to crack the coconut non-social is not that it is unopposable, but instead that it does not affect others significantly. Perhaps we should say simply:

Third Definition

My power to effect O is social power iff

(iii) O affects others significantly.

This would follow Steven Lukes’ classic Power: A Radical View:

The absolutely basic common core to, or primitive notion lying behind, all talk of power is the notion that A in some way affects B. But in applying that primitive (causal) notion to the analysis of social life, something further is needed – namely, the notion that A does so in a non-trivial or significant manner (Lukes, 2021: 35).

What is it to affect others significantly? One possibility would be to give a list. It is to affect significantly others’ internal states (beliefs, preferences, etc.), actions, external opportunities, cost structure and so on. I suspect that when one asks, first, what unifies that list and, second, what counts as significantly affecting the items on the list, one will have to have recourse to others’ interests.

This suggests a second possibility, namely, that to affect others significantly is to affect their interests significantly. This would again follow Lukes, who goes on to say that the ‘underlying concept of power’ is one ‘according to which A exercises power over B when A affects B in a manner contrary to B’s interests’, at least if we take into account Lukes’s (2021: 69–70) later acknowledgement that A can also affect B’s interests by promoting B’s interests. There is a question of how ‘interests’ is to be understood. Although not much turns on it for the purposes of this article, I agree with Lukes that ‘interests’ must be understood objectively, rather than subjectively, such as in terms of the satisfaction of desires or preferences. Among other reasons for adopting an objective understanding of interests is that we might otherwise fail to count as social power the power to modify desires or preferences themselves.

To be sure, ‘significantly’ is vague. On Third Definition, ‘social power’ inherits its vagueness. However, it is not clear why this should be a problem. The concept of social power may simply be vague.

9

Against Third Definition, intuitively our example of omnilateral power – unanimity voting – would still be a case of social power even if O did not affect others significantly. Likewise, intuitively our example of unilateral power – Two Votes – would still be a case of social power even if O did not affect others significantly. This suggests:

Fourth Definition

My power to effect O is social power iff

(i) Someone can assist me in effecting O

I.e., there is a C in which I have a m to O such that if I m, I effect O only if and partly because for some X, a, X a’s

or

(ii’) Someone can resist my effecting O, and if they resist, I can overcome their resistance.

I.e., there is a C in which I have a m to O and some Y has a way of resisting, r, my effecting O, such that if I m, Y r’s, and other agents n, then I still effect O.

or

(iii) O affects others significantly.

This means that dictator-rule voting, an example of unopposable power, counts as social when but only when O affects others significantly. This seems like the right thing to say. If O is that a coconut is cracked, which does not affect others significantly, then dictator-rule voting about whether to crack the coconut, with me as the dictator, is no different from my being able to crack it forthwith on the desert isle.

10

Do we need all three disjuncts of Fourth Definition, however? First, (iii) can hold without (i) and (ii’). This is illustrated by dictator rule voting where O affects others significantly. Second, (i) can hold without (ii’) or (iii). This is illustrated by unanimity voting where O does not affect others. Third, (ii’) can hold without (iii). This is illustrated by Two Votes where O does not affect others.

Finally, however, (ii’) cannot hold without (i). Informally and roughly, if (ii’) holds, then someone can resist me, and if someone can resist me, then someone assist me by not resisting me, and if someone can assist me, then (i) holds. For a formal argument, see steps (4)–(7) in the argument in section 4.

For an illustration, consider Two Votes. Since you can resist by casting your vote against O, then you can also assist by not casting your vote against O. Objection:

Suppose that you will cast your vote against O no matter what. (Suppose, for example, we wish to ask, given that you will vote against O, what sort of power, if any, do I have to effect O?) Then you can’t assist by not casting your vote against O.

Reply: By hypothesis, it cannot be the case that you vote against O no matter what. If you vote against O no matter what, then my casting only one vote for O is no longer a means to O; there is no longer a pattern of others’ actions such that if I cast only one vote I effect O. If my casting only one vote for O is not a means to O, then your voting against O does not count as resistance. (Indeed, and somewhat paradoxically, you have made my power unopposable; my only means to O is to cast both votes, which effects O no matter what you do.) But, by hypothesis, your voting against O does count as resistance.

If (ii’) implies (i), then (ii’) can drop out of our definition. Thus, we have:

Final Definition

My power to effect O is social power iff

(i) Someone can assist me in effecting O

I.e., there is a C in which I have a m to O such that I effect O partly because I m only if and partly because for some X, a, X a’s

Or

(ii) O affects others significantly.

In short, social power is power that either has others as partners or has others as patients.

Abizadeh argues persuasively that not all social power is power to overcome resistance. But one can go further. Talk of overcoming resistance can drop out of the definition of social power altogether. The grammar of social power need not have a place for a term involving overcoming resistance.

11

Note that power to affect others significantly need not be ‘power-over’ them, as Abizadeh (2021a: 4) understands it. ‘The concept of power-over, as I construe it, refers to a capacity to effect specific states, dispositions, or actions in other agents (patients) in conformity with the power-holder’s will or interests’. Later, he clarifies that the ‘states, dispositions, or actions’ are ‘internal’, by which he seems to mean ‘mental’ or ‘psychological’ (Abizadeh, 2021a: 5).

I might have the power to kill you, lame you, expose you to disease, confine you, deprive you of food or water and so on, all of which affects you significantly, without the power to effect specific changes in your psychology in conformity with my will or interests. So it is not only disjunct (i) that vindicates Abizadeh’s claim that social power need not be power-over, as witnessed, say, by the example of unanimity voting on a question that does not affect others. It is also disjunct (ii) that vindicates Abizadeh’s claim that social power need not be power-over, as witnessed by power of mine that significantly affects you without effecting specific changes in your psychology in conformity with my will or interests.