236 Chapter 9

implement (3,3). If the Soviets possess threat power in game 30, the
outcome would not change, because their compellent threat of choosing
W would also induce (3,3).
Thus, who possesses threat power is irrelevant in this game. Although
irrelevant, however, its impact is certainly salutary in allowing the players
to avoid (2,2) and obtain (3,3) instead. This is significant in game 30,
because neither player has a dominant strategy, there are no pure-
strategy Nash equilibria, and, consequently, there is no indubitably
rational choice according to standard game theory.
The two different representations of the Cuban missile crisis testify to
the need to ponder strategic conflicts from different perspectives that
take into account the perceptions of players, or possibly misperceptions,
as in the Iran hostage crisis (section 9.4). In the fashion of Rashomon
(a Japanese movie that portrays four different versions of a rape), each
perspective gives new insights. It is especially instructive to see how
sensitive outcomes are to the different reconstructions on which each is
based and to consider the relationship of these to the actual outcome. In
section 9.6, I explore the possibility of deception in the Cuban missile
crisis.

9.6 Deception in the Cuban Missile Crisis

Defi ne a player’s deception strategy to be a false announcement of its

preferences to induce the other player to choose a strategy favorable to
itself (i.e., the deceiver).20 I assume that the deceived

1. has no information about the deceiver’s true preference ranking.21

The deceived, therefore, has no basis for mistrusting the deceiver. I also
assume that the deceived

2. does not have a dominant strategy.

20. Game-theoretic models of deception, including some with applications, can be found
in Brams 1977; Brams and Zagare 1977, 1981; Zagare 1979; Muzzio 1982; and Board 2002.
A useful compilation of material on deception, both in theory and practice, is Daniel and
Herbig 1982.
21. This was not true in the Iran hostage crisis. Although Carter misperceived Khomeini’s
preferences, he was correct about Khomeini’s most-preferred state (4), namely NO (figure
9.2).
Incomplete Information in Literature and History 237

Otherwise, the deceived would always choose it, regardless of what the
deceiver announced as its own strategy.
A deception strategy, like a threat strategy, requires prior communica-
tion, but to indicate preferences rather than threaten some action. Given
such communication and that conditions 1 and 2 are met, the deceiver,
by pretending to have a dominant strategy, can induce the deceived to
believe that it will always be chosen. Anticipating this choice, the
deceived will then be motivated to choose its strategy that leads to the
better of the two states associated with the deceiver’s (presumed) domi-
nant strategy.
In the case of deception, I assume that play starts with the communica-
tion of (false) preferences by the deceiver. The players choose strategies,
but play does not commence from a particular state. I will return to this
point later and discuss its consistency with TOM, but first I consider the
possible use of deception in the Cuban missile crisis.
As the crisis heightened, the Soviets indicated an increasing predispo-
sition to withdraw rather than maintain their missiles if the United States
would pledge not to attack Cuba and not to invade it in the future. In
support of this shift in preferences, contrast two statements by Premier
Khrushchev, the first in a letter to the British pacifist, Bertrand Russell,
the second in a letter to President Kennedy:

If the way to the aggressive policy of the American Government is not blocked,
the people of the United States and other nations will have to pay with millions
of lives for this policy. (Divine 1971, 38)

If assurances were given that the President of the United States would not par-
ticipate in an attack on Cuba and the blockade lifted, then the question of the
removal or destruction of the missile sites in Cuba would then be an entirely
different question. (Divine 1971, 47)

Finally, in an almost complete about-face, Khrushchev, in a second letter

to Kennedy, all but reversed his original position and agreed to remove
the missiles from Cuba, though he demanded a quid pro quo (which was
ignored by Kennedy in his response, quoted in section 9.5):
We agree to remove those weapons from Cuba which you regard as offensive
weapons. . . . The United States, on its part, bearing in mind the anxiety and
concern of the Soviet state, will evacuate its analogous weapons from Turkey.
(Divine 1971, 47)
238 Chapter 9

Khrushchev, who had previously warned (in his first letter to Kennedy)
that “if people do not show wisdom, then in the final analysis they will
come to clash, like blind moles” (Divine 1971, 47)—which I suggested
earlier (chapter 7, note 21) indicated his sense of loss of control—seemed,
over the course of the crisis, quite ready to soften his original position.
This is not to say that his later statement misrepresented his true pref-
erences—on the contrary, his language evoking the fear of nuclear war
has the ring of truth to it. Whether he actually changed his preferences
or simply retreated strategically from his earlier pronouncements, there
was a perceptible shift from a noncooperative position (maintain the
missiles regardless) to a conditionally cooperative position (withdraw
the missiles if the United States would also cooperate).
Perhaps the most compelling explanation for Khrushchev’s modifica-
tion of his position is that there was, to use Howard’s (1971, 148, 199–201)
apt expression, “deterioration” in his original preferences in the face of
their possibly apocalyptic consequences if they were acted upon. By
interchanging, in effect, 3 and 4 in the Soviet ranking of states in the
figure 9.4 representation of the crisis (game 30), Khrushchev made W
appear dominant, thereby inducing the United States to cooperate
(choose B). The resulting (3,4) state is next-best for the United States,
best for the Soviet Union, and renders BW a Nash equilibrium in this
putative game.22
Whether Khrushchev deceived Kennedy or actually changed his pref-
erences, the effect is the same in inducing the compromise selected by
both sides. Although there seems to be no evidence that conclusively
establishes whether Khrushchev’s shift was honest or deceptive, this
question is not crucial to the analysis. True, I have developed the analysis
in terms of rational deception strategies, but it could as well be inter-
preted in terms of a genuine change in Khrushchev’s preferences,
given that preferences are not considered immutable.
Could the United States have deceived the Soviets to induce (3,3) in
game 30? The answer is no: If the United States had made B appear
dominant, the Soviets would have chosen M, resulting in (1,4); if
the United States had made A appear dominant, the Soviets would
have chosen W, resulting in (2,2). Paradoxically, because the United

22. This game is 21 in the appendix, and (3,4) is also the unique NME in this game.
Incomplete Information in Literature and History 239

States, as a deceiver, could not ensure an outcome better than its

next worst (2)—whatever preference it announced—it was in its interest
to be deceived (or at least induced) in order that (3,3) could be
implemented.
More generally, in five of the fifty-seven 2 × 2 conflict games (29, 30,
3l, 46, 47), at least one player can do better as the deceived than the
deceiver, in terms of a player’s comparative rankings of outcomes. Thus,
it may be profitable that the deceived not know the preferences of the
deceiver, and for the deceiver to know that the deceived does not know,
and so on ad infinitum. For this set of five games, the strange notion that
“ignorance is strength” seems well founded.23
Now consider the consequences for both sides if they had “played it
safe” in the game 30 representation of the Cuban missile crisis by choos-
ing their “security-level strategies.” (A player’s security level is the best
payoff that it can ensure for itself, whatever contingency occurs, which
in this case is the United States’ next-worst payoff of 2; a player’s pure
strategy associated with this payoff is its security-level strategy.) The
choice of such a strategy to avoid its worst state (1) means the United
States would choose A; if the Soviets also choose their security-level
strategy (W), the resulting state is (2,2), which is Pareto-inferior to (3,3).
In section 9.5 I argued that not only is (3,3) the unique NME in game
30 but also that it can be induced by the moving power of either player,
by a deterrent threat of the United States, and by a compellent threat of
the Soviet Union. Thus, (3,3) is a compelling solution in game 30, even
though this game is strongly cyclic and has (2,2) as its security-level
outcome.
Khrushchev had good reason to try to enhance (3,3)’s attractiveness
to the United States by making it appear that he definitely would choose
W because of its dominance. By seeming to interchange 3 and 4 in his

23. These five games are a subset of the “deception-vulnerable” games, which comprise a
total of seventeen of the seventy-eight 2 × 2 strict ordinal games (22 percent); twenty-one
games (27 percent) are “deception-proof,” and forty games (51 percent) are “deception-
stable” (Brams 1977). Deception-vulnerable games are the only games in which deception
is both possible—because both players do not have dominant strategies (unlike deception-
proof games)—and profi table—because at least one player can induce a better outcome
by announcing a false preference order (unlike deception-stable games). All the 2 × 2
games that fall into these mutually exclusive categories, as well as various subcategories,
are given in Brams 1977.
240 Chapter 9

preference ordering, he transformed game 30 into game 21 (see appen-

dix), which entirely robs him of any incentive to depart from W.
This explanation for Khrushchev’s about-face, grounded in standard
game theory, makes a good deal of sense. But there is an alternative
explanation, based on TOM, which seems to me equally plausible.
At the start of the crisis, the state was (1,4) in game 30: The Soviets
were in the process of installing their missiles, and the United States
announced that it would blockade further shipments. According to TOM,
the migration from (1,4) to the NME of (3,3) in game 30 must proceed
through the two other states, (4,1) and (2,2), as shown in figure 9.4. By
contrast, in game 21, it is rational for the Soviets to move directly from
(1,3) to the NME of (3,4).
Thus, the compromise state is, in a sense, more efficiently achieved in
game 21 than game 30, according to TOM. This efficiency-based explana-
tion of why Khrushchev sought to change the game offers an alternative
way of viewing, within the TOM framework, the rationality of deception
in this crisis.
I will not analyze the efficiency of moves in games generally. But it is
certainly related to the feasibility of moves discussed in section 2.6,
where I argued that certain moves might be impossible. Even when
moves are possible, there may be more than one path to an NME,
in which case players will presumably prefer to make fewer moves to
reach it.
Changing the appearance of a game through deception may contribute
to this greater efficiency. When it does, it is reasonable to suppose that
players will use deception, especially if it can go undetected in games of
incomplete information.24

9.7 The Paradox of Omniscience

Hamlet well illustrates a game in which information is at a premium.

Throughout the play, each of the antagonists tries to learn more about

24. The game-theoretic models cited in note 20 distinguish between “tacit” deception,
which is undetected, and “revealed” deception, which is detected. This distinction is illus-
trated by the 1954 Geneva negotiations over the future of the two Vietnams, which involved
three players and are analyzed in Zagare 1979; see also Brams 1990, chap. 8.