Ethical issues in moral and social enhancement

7.6.3. Arbitrary and unchecked power 
The understanding of what arbitrary power consists of merits some explication. 
There are two broad approaches to that question. The first approach is to define 
‘arbitrariness’ substantively. According to this approach, power is arbitrary when it 
fails to track what Pettit called the ‘welfare and world-view’ (1997, p. 56) of those 
affected. This substantive definition is open to at least three interpretations, 
depending on whether we interpret the ‘welfare and world-view’ of those affected 
as a) their objectively-defined interests, b) their subjective preferences, or c) their 
shared ideas as expressed through an appropriate deliberative process. This 
approach has sparked a complex discussion on the difference between freedom and 
the common good and the possibility of collapsing them into each other, which I 
will not examine here (see for example: Larmore, 2004; Costa, 2007; Carter, 2008 
for discussion). For now it suffices to say that in the God Machine scenario as 
stated by 
Savulescu and Persson (2012a),
there is no obvious reason to think that 
those conditions are not fulfilled.
The second broad approach is procedural. To move away from the difficulties 
brought on by the discussion of the definition of the ‘arbitrary’ in relation to the 
common good, Pettit has proposed abandoning the notion of arbitrary power and 
focusing instead on whether the power is controlled or uncontrolled 
(2012, p. 58). 
On this view, power is arbitrary or unchecked to the extent that it is not reliably 
constrained by effective rules, procedures, or goals that are common knowledge to 
all persons or groups concerned (Lovett, 2001, 2010). The appropriate ‘check’ may 
come in the shape of control by the persons specifically affected (the democratic 
view), or control by the society’s laws, norms and institutions (the procedural 
view).

---

166 
According to Lovett (2012), it is important that the constraints put on the exercise 
of power are ‘reliably effective.’ In order to be reliably effective, constraints must 
remain effective over a wide range of possible changes or modifications in the 
relevant circumstances. In a similar context, Pettit proposed ‘non-manipulability’ as 
a criterion institutional ‘instruments’ should satisfy:
‘Designed to further certain public ends, they should be 
maximally resistant to being deployed on an arbitrary, perhaps 
sectional, basis. No one individual or group should have discretion 
in how the instruments are used... The institutions and initiatives 
involved should not allow of manipulation at anyone’s individual 
whim.’ (Pettit, 1997, p. 173) 
He then lays out three conditions that a ‘non-manipulable’ system must satisfy. 
This includes firstly the rule of law, according to which laws ‘should be general and 
apply to everyone, including the legislators themselves; they should be promulgated 
and made known in advance to those to whom they apply; they should be 
intelligible, consistent, and not subject to constant change; and so on’ (Pettit, 1997, 
p. 174). Secondly, the dispersion-of-power condition requires that ‘powers which 
officials have under any regime of law should be dispersed’ by mechanisms such as 
the separation of powers, bi-cameralism, federalism, and international legalism 
(Pettit, 1997, pp. 177–80). The third condition states that laws should be insulated 
from ‘excessively easy majoritarian change’ (Pettit, 1997, p. 180).
Does the God Machine satisfy these conditions? There is no reason why it should 
not be possible. Taking Persson and Savulescu’s (2012a) scenario at face value, 
there is no reason to suspect that the God Machine society would necessarily fail to 
institute such protections. Even systems that do what is beyond the ability of one 
human to do can be reviewed and controlled – for example, the assumptions on 
which the computations were made can be accessible to further analysis by 
computer systems independent of the God Machine, predictions generated by the 
God Machine can be tested and the system’s workings subject to regular audit by a 
group of auditors immune to the machine’s influence, every intervention of the God 
Machine might have to be reviewed by an independent computer system and a 

---

167 
human, etc.
36
The second question would be about whether the God Machine is 
vulnerable to abuse by others. This question can only be answered with reference to 
the political context that the God Machine is placed in. To the extent that this is 
presented by the authors, one can suspect that in a morally (and politically?) 
enhanced society appropriate checks on the access to the God Machine will be 
placed, and the representative system characterised by legitimacy and transparency 
is likely to be stable.
The most convincing argument in this context (Harris, 2014a) comes from the fact 
that the God Machine society’s citizens would not know when the lawful 
intervention would take effect, and thus have no opportunity to question, respond to 
or challenge this intervention. The God Machine would therefore lack an important 
feature that allows for the protection of liberty. Overall however, if we take the God 
Machine at face value and thus accept the facts about the possible world it was 
placed in, the arguments that attempt to establish that the God Machine powers are 
beyond societies’ review largely fails. 
The God Machine is not necessarily inimical to appropriate control, at least as 
presented by 
Savulescu and Persson (2012a). 
As a result, Sparrow’s account does 
not support either plausible interpretation of freedom from domination as the right 
kind of structural relation between the God Machine and the citizens – the God 
Machine does not obviously fail the ‘checks and balances’ requirement and it is not 
clear that the God Machine would serve to establish a relation of domination 
between some members of the God Machine society and others. A further argument 
may develop the in-principle problems of auditing and controlling an entity like the 
God Machine, but in the absence of such an argument, the objection fails.