In this, the latest and most comprehensive word on Pascal's Wager, Jeff Jordan defends the rationality of one pragmatic argument for theism. He scores some success, although important qualifications must be made.

Chapter 1 identifies a variety of Pascalian arguments for theism, emphasizing the Canonical Wager and the Jamesian Wager. The Canonical Wager is the familiar ‘argument from dominating expectation’, as reported in Ian Hacking's contribution to Jordan's excellent essay collection, Gambling on God (Lanham MD: Rowman & Littlefield, 2000). According to this version, even a remote possibility of heaven means that theism enjoys infinite expected utility; the expected utility of naturalistic atheism is only finite; therefore you should believe in God (23). According to the wager as developed by William James, one should believe in God because theists are better off than atheists during their lives on earth. Jordan ultimately disowns the Canonical Wager and embraces the Jamesian Wager.

Chapter 2 rejects ‘absolute evidentialism’, which requires the suspension of judgement when the evidence is equivocal, in favour of ‘defeasible evidentialism’, which calls for respecting evidence only insofar as it speaks. More precisely, absolute evidentialism makes the absolute and biconditional claim that S ought to believe P if and only if S's evidence supports P; defeasible evidentialism is the hedged conditional that S ought to believe P if S's evidence supports P, unless S is permitted otherwise (45). Against absolute evidentialism, Jordan considers an Alpine hiker whose only real hope of survival is to jump a chasm, and whose best hope at successfully jumping lies in convincing himself that he can. The conclusion ‘is that pragmatic belief-formation is sometimes both morally and intellectually justifiable’ (47).

Jordan's defeasible evidentialism, as I read it, is equivalent to a logical truth: if you ought to believe P, then if your evidence supports P, you ought to believe P. If so, that undermines the interest of Jordan's claim that Pascalians, ‘contrary to conventional thought, can be evidentialists’ (53). Incidentally, Jordan adds a curious qualification: ‘if there is a belief-inducing technology that works even when the known evidence is extremely unfavorable, a commitment to Defeasible Evidentialism … precludes employing it – no matter how beneficial the falsehood’ (196). It's not clear why this should be. If leaping a chasm is truly your only hope of survival then, all else equal, even if there is only a one in a million chance that deceiving yourself about your chasm-leaping capacities would do any good, it would seem to be justifiable (morally, not epistemically).

Chapter 3 responds to the many-gods objection, which comes in two main varieties. According to the possibilist objection, there are countless possible gods, some of whom would punish standard theists; consequently expected utilities for various religious beliefs, being either incalculable or evenly balanced, neutralize each other. According to the actualist objection, expected utilities are neutralized by just ‘live options’ alone.

Against the possibilist objection, Jordan writes:

Calling a philosophical hypothesis a fiction, needless to say, begs the question. More significantly, I am not convinced that Jordan overcomes the rebuttals raised in earlier work of mine (‘Pascal's Wager and the many gods objection’, Religious Studies, 37 (2001), 321–341). For example, cockroach-gods strike me as no more fantastic than anthropomorphic gods for several reasons, one being the existence of human suffering. Because Jordan is right to think that ‘the experiences and reflections of our community, our intellectual peers, must be accorded some epistemic weight’ (81), and because Jordan and I are peers, and because Jordan knows I've insisted on assigning some very small probability to the cockroach-god, it seems to me that Jordan too must assign some epistemic possibility to the cockroach-god (perhaps discounted, but still non-zero).

Perhaps it is because of the many-gods objection that Jordan moves to the Jamesian Wager. According to William James, religious belief carries great utility in this world. It produces ‘a new zest which adds itself like a gift to life, and takes the form either of lyrical enchantment or of appeal to earnestness and heroism … an assurance of safety and a temper of peace and, in relation to others, a preponderance of loving affections’ (89). Quoting empirical studies, Jordan adds that religion is a ‘source of better mental health and even physical health’ and is ‘the best predictor of life satisfaction’ (90). These claims, inevitably, will face challenges, but Jordan anticipates them: unless one knew one way or the other, he writes, ‘the prudential response is to proceed as if the religious commitment produces the benefits’ (94).

Chapter 4 responds to problems arising from infinite utilities. As first noted by Antony Duff (‘Pascal's Wager and infinite utilities’, Analysis, 46 (1986), 107–109), any decision of mine whatsoever might initiate a chain of events somehow leading to my salvation. If deciding to believe in God gives me a 0.5 chance at salvation, while deciding to reject God gives me a 0.001 chance, then the expected utilities of believing and disbelieving are the same (0.5×infinity=infinity, and 0.001 x infinity=infinity). Jordan responds to this problem by positing an addendum to standard decision theory, principle N: ‘[I]f each available action has infinite expected utility then, all else equal, perform that action that is most likely to bring about the pay-off’ (104). Jordan believes N is ‘plausible’, although some would disagree.

Despite his reply to Duff (and his reply to the St Petersburg paradox, to which I have similar reservations), Jordan recognizes that the use of infinite utilities is jointly inconsistent with the monotonicity and expectation rules of standard decision theory (119), and he advises the Pascalian ‘to jettison the idea of an infinite utility’ (120). In that case, it is important to consider the specific probabilities of rival metaphysical hypotheses. For ‘deep atheists,’ those who believe that the probability of God's existing is vastly under 0.5, the Wager has no force (122).

Chapter 5 runs through a number of remaining objections, for instance that Pascalian reasoning is mercenary, that it conflicts with predestinationism and the divine plan, and that it succumbs to criticisms by Hajek, Oppy, Swinburne, and others.

Chapter 6 begins by ‘situating the Wager within the Anselmian project’, which holds that evidence for theism can actually be discerned if one is just willing to believe (for Pascal, atheism can result from ‘only concupiscence and malice of heart’ (172)). As a result, Jordan sees the Wager as a bridge from pragmatic justification to epistemic justification, and as being compatible with absolute evidentialism after all (173). Chapter 6 goes on to discuss pragmatic arguments other than Pascal's Wager. Jordan endorses J. S. Mill's view that, in the absence of evidence either way, one is entitled to hope for life after death, even if one is not entitled to believe in it (187–190); he endorses the argument of William James, that hope legitimately factors into one's decisions as to what to believe (174–186); and he says that ‘one hopes in the Christian sense only if one believes in God’ (198).

It is important for Jordan's Pascalian project, and indeed for his larger theistic project, that anti-theistic arguments fail. Chapter 7, the last, accordingly attempts to undermine the divine hiddenness argument: that (a) if there is a very powerful loving God then God would clearly reveal Himself to me; (b) God has not clearly revealed Himself to me; therefore (c) there is no powerful loving God. Jordan raises standard alternatives to (a), for instance that evidence of God ‘is a divine gift of which justice precludes universal distribution’ (203). This is supposed to establish that (c) is not a necessary truth. However, can it not remain that (a), and therefore (c), are very probable? For many, verbal formulations of a loving yet inequitable and secretive God are as possible as an earthly father who claims to be good yet never allows a subset of his children to know of his existence, even though they want to. To judge God by different lights than we would judge anyone else sounds like special pleading.

Jordan also attempts to undermine premise (b) by observing that Pascal's Wager itself might count as a reason God has given us for theism. This is an interesting move, but I remain unconvinced. The soundness of pragmatic reasoning in favour of theism is unclear, as attested by its long controversial history. It is unclear not only to those who have found it wanting, it is unclear to those, both prior to Pascal's day and subsequently, who have never thought of it.

Jordan claims that my work, cited above, confuses atheism with non-theism (meaning, for Jordan, ‘no belief in supernatural reality’ and ‘no belief in supernatural person’). I stand by my word choice, and more importantly I stand by my intended claim: that some version of Buddhism, regardless of whether it be called atheistic or non-theistic, undermines the Pascalian reason for believing in God. Jordan appears to accept this when he replies that ‘Religious belief of some sort is what rationality demands … . Although the Ecumenical Wager may not support theism as the only rational option, it still plays a vital role of undermining the rationality of the naturalistic options.’ When Jordan also states, ‘that theistic belief carries a greater expected utility than does disbelief, and so one ought to believe’ (86, italics added), he means that religious belief carries a greater expected utility, and that theism carries the day in those societies where theistic religions are associated with fewer conversion costs than atheistic religions (personal communication). It should also be noted that Pascal's Wager, for Jordan, is sometimes an extended family that includes the Jamesian Wager, but sometimes, more narrowly, it is distinct from the Jamesian Wager. These one or two points aside, however, Jordan's exposition is admirably clear.

To summarize, Jordan argues that it is rational for a subject, S, to subscribe to theism so long as the following conditions are satisfied: (1) S's evidence and reasoning must not weigh too heavily against theism; (2) S must be closed off to infinitely many ‘philosopher's fictions’, yet open to the kinds of god who reward Pascalian wagering; (3) the expected earthly utility for S's maintaining theism must outweigh that of all available alternatives; and (4) theism is to be understood as religious belief quite generally, not necessarily belief in a being who hears our prayers.

When conditions (1–4) are satisfied, subjects may indeed think that it is prudentially rational to maintain theism, and in this respect Jordan's book will succeed in encouraging many who are theists already. However, I question how many agnostic readers will satisfy condition (2), and I wonder how much is accomplished if those agnostics disposed to follow conventional thinking, as represented by satisfaction of (2), drift into diffuse, non-theistic religiosity, which is all that Jordan's Wager calls for. More fundamentally, in order truly to be rational, subjects must have reason to reject ‘philosophers' fictions’ – simply claiming to know that they are false, and calling them bizarre, does not count as a reason (this general point is pressed by Craig Duncan, ‘Do vague probabilities really scotch Pascal's Wager?’, Philosophical Studies, 112 (2002) 279–290). At the same time, I wonder whether (2) is really necessary if (3) holds. Be that as it may, Jordan's sophisticated treatment of Pascal's Wager is packed with original and provocative contributions to decision theory, evidentialism, and the Jamesian Wager. It is a must-read for everyone in these areas, and it would also make a splendid text for upper-division and graduate-level courses.