Some representation problems for semiorders

Suppose we have a number representation of a semiorder 〈A, P〉 such that aPb iff f(a)+δ(a) < f(b), for all a, b ∈ A, where δ is a nonnegative function describing the variable jnd. Such an f (here called a closed representation) may not preserve the simple order relation $\text{R}{}^1$ generated by 〈A, P〉, i.e., $\text{aR}{}^1\text{b}$ but f(a) > f(b) for some f, δ and a, b ∈ A. We show that this “paradox” can be eliminated for closed and closed interval representations. For interval representations it appears to be impossible. That is why we introduce a new type of representation (an R-representation) which is of the most general form for number representations that preserve the linear structure of the represented semiorders. The necessary and sufficient condition for an R-representation is given. We also give some independent results on the semiorder structure. Theorems are proved for semiorders of arbitrary cardinality. The Axiom of Choice is used in the proofs.     

Two-generator semigroups of binary relations

We study upper bounds on the size of the semigroups generated two randomly chosen n × n Boolean matrices having exactly N one entries. In general, as $\text{N}\text{n}$ increases the semigroups tend to be smaller. If N is unrestricted, or is a function which is at least [$\text{((2 + ?)n}{}^3\text{log}\text{n)}{}^{\mathrm{<mtext>1</mtext><mtext>2</mtext>}}$] the probability tends to one that the size of the semigroup is exactly 3. If N is a function which is at least [(r + 1 + ?)n log n] the probability tends to one that the size of the semigroup will be no more than $\text{2}{}^{\mathrm{<mtext>(</mtext><mtext>(n?1)</mtext><mtext>r</mtext><mtext>)+1</mtext>}}\text{? 1}$. However, if N is any function w(n) such that $\text{w(n)}\text{n}{}^2\text{→ 0}$ and w(n) > n, the average size of the resulting semigroups will be at least . This phenomenon is caused by some semigroups of extremely large size. It is thought that w(n) > n and $\text{w(n)}\text{n}{}^2\text{→ 0}$ hold for those matrices usually encountered in sociology and psychology.     

On the functional form of partial masking functions in psychoacoustics

Partial masking of pure tones is often investigated in terms of matching functions which record the intensity φ(x, n) of an unmasked tone which matches, in perceived loudness, a tone of intensity x embedded in a flat, broadband noise of intensity n. Empirically, the following property of “shift invariance” is observed to hold for these matching functions: φ(λx, λ^θn) = λφ(x, n), for any λ > 0, and some θ < 1. In the context of the model $\text{φ(x, n) = F}{}_0\text{[}\text{g}{}_0\text{(x)}\text{(h}{}_0\text{(x) + k}{}_0\text{(n))}\text{]}$ which expresses the idea that the effect of a noise mask is to control the gain of a pure tone signal in a multiplicative fashion, shift invariance proves to be an extremely powerful theoretical constraint. Specifically, we show that only two parametric families of functions (F₀, g₀, h₀, k₀) are possible candidates for interpolating empirical data. These two parametric families are given by the following expressions: $\text{φ}{}_1\text{(x, n) = A(}\text{x}{}^{\mathrm{\alpha }}\text{(x}{}^{\mathrm{\alpha \prime }}\text{+ Kn}{}^{\mathrm{<mtext>\alpha \prime </mtext><mtext>\theta </mtext>}}\text{)}\text{)}{}^{\mathrm{<mtext>1</mtext><mtext>(\alpha \; ?\; \alpha \prime )</mtext>}}$, $\text{φ}{}_2\text{(x, n) = A[x}{}^{\mathrm{\alpha }}\text{(x}{}^{\mathrm{\alpha \prime }}\text{? Kn}{}^{\mathrm{<mtext>\alpha \prime </mtext><mtext>\theta </mtext>}}\text{)]}{}^{\mathrm{<mtext>1</mtext><mtext>(\alpha \; +\; \alpha \prime )</mtext>}}$. Both of these expressions are in good agreement with a large array of partial masking data.     

Conjoint Weber laws and additivity

This paper investigates the mathematical consequences of a number of related empirical laws, exemplified by

$\text{P}{}_{\mathrm{ax;by}}\text{= P(ξa)(ξx);(ξb)(ξy)}$

where a, x, b, y, and ξ are real numbers, and P_ax;by is the probability of choosing the two-dimensional object (a, x) in the set {(a, x), (b, y)}. A variety of results is derived showing that, in the presence of such laws, the class of feasible models for choice data is considerably reduced. In particular, it is shown that the above law, together with the “additive conjoint” form

$\text{P}{}_{\mathrm{ax;by}}\text{= F[l(a) + r(x), l(b) + r(y)]}$

(where F, l, and r are unspecified except for continuity and monotonicity properties), requires the choice probabilities to possess one of the following three analytic forms:

$\text{P}{}_{\mathrm{ax;by}}\text{= G}\text{a}{}^{\mathrm{\beta }}\text{+ δx}{}^{\mathrm{\beta }}\text{b}{}^{\mathrm{\beta }}\text{+ δy}{}^{\mathrm{\beta }}\text{, β ≠ 0}$

;

$\text{P}{}_{\mathrm{ax;by}}\text{= G(a}{}^{\mathrm{\beta }}\text{x}{}^{\mathrm{\gamma }}\text{/b}{}^{\mathrm{\beta }}\text{y}{}^{\mathrm{\gamma }}\text{), β + γ ≠ 0}$

;

$\text{P}{}_{\mathrm{ax;by}}\text{= Q}{}_0\text{(a/x, b/y)}$

.     

Conditions under which thurstone case III representations for binary choice probabilities are also Fechnerian

By a Thurstone Case III representation for binary symmetric choice probabilities P_x,y we mean that there exist functions F, μ, σ > 0 such that $\text{P}{}_{\mathrm{x,y}}\text{= F[}\text{(μ(x) ? μ(y))}\text{(σ}{}^2\text{(x) + σ}{}^2\text{(y))}{}^{\mathrm{<mtext>1</mtext><mtext>2</mtext>}}\text{]}$. We show that the constraint σ = constant, or μ = ασ + β, α ≠ 0, is both necessary and sufficient for a Thurstone Case III representation to be Fechnerian, i.e., to be reexpressable as as P_x,y = G(u(x) ? u(y)) for some suitably chosen functions G, u.     

Binary choice probabilities: on the varieties of stochastic transitivity

Let {P_λ} denote the family of decisiveness relations {$\text{P}{}_{\mathrm{\lambda }}\text{:}\text{1}\text{2}\text{≤ λ < 1}$} with aP_λb if and only if P(a,b) > λ, where P is a binary choice probability function. Families in which all decisiveness relations are of the same type, such as all strict partial orders or all semiorders, are characterized by stochastic transitivity conditions. The conditions used for this purpose differ in various ways from the traditional forms of strong, moderate, and weak stochastic transitivity. The family {P_λ} is then examined from the viewpoint of interval representation models, the most general of which is aP_λb if and only if I(a, λ) > I(b, λ), where the I's are real intervals with I(a, λ) > I(b, λ) if and only if the first interval is completely to the right of the second. With I(a, λ) = [f(a, λ), f(a, λ) + σ(a, λ)], the specializations of the interval model that are discussed include those where the location function f (for left end-points) depends only on the set A of alternatives or stimuli and where the length function σ depends only on A or on λ or neither.     

On the reciprocity of proximity relations

The degree of reciprocity of a proximity order is the proportion, P(1), of elements for which the closest neighbor relation is symmetric, and the R value of each element is its rank in the proximity order from its closest neighbor. Assuming a random sampling of points, we show that Euclidean n-spaces produce a very high degree of reciprocity, $\text{P(1) ≥}\text{1}\text{2}$, and correspondingly low R values, E(R) ≤ 2, for all n. The same bounds also apply to homogeneous graphs, in which the same number of edges meet at every node. Much less reciprocity and higher R values, however, can be attained in finite tree models and in the contrast model in which the “distance” between objects is a linear function of the numbers of their common and distinctive features.     

On unpredictability as a causal factor in “learned helplessness”

In two replications, two groups of dogs were exposed to a series of uncontrollable, electric shocks. For one group the shocks were preceded by a tone (i.e., Paired). For the other group the shocks were randomly related to the tones and hence unpredictable (i.e., Random). Each replication also included a third group; in the first it was exposed only to the series of tones (CS-only), while in the second, it was exposed only to a series of shocks (Shocks-only). Then, all dogs were required to learn a discriminative choice escape/avoidance task in which the required response was to lift the correct paw in the presence of each of two visual S^Ds to escape or avoid the shocks [($\text{S}{}_1{}^{\mathrm{D}}\text{?R}{}_1\text{)}\text{(S}{}_2{}^{\mathrm{D}}\text{?R}{}_2$)]. Dogs preexposed to random tones and shocks were least successful in learning the task relative to those groups which experienced either predicted shocks, only the tones, or only the shocks, which in turn did not differ from each other. These results permitted the inference that the proactive interference with choice behavior following random tone CSs and shocks was attributable to a learned irrelevance generalized with respect to CSs.     

Decision making in a case of mixed uncertainty: A normative model

We consider a case of uncertainty which is frequently met in various fields, e.g., in parametric statistics: Events {θ}, θ ∈ ∵, are members of family $\text{E}$ on which the decision maker possesses no information at all; however, conditionally on the realization of {θ}, he is able to affix probabilities to all members of another family of events, $\text{F}$. We assume that the decision maker: (1) has a rational behavior under complete ignorance, for decisions whose results only depend on events of $\text{E}$; (2) with {θ} known, maximizes his conditional expected utility for decisions whose results only depend on events of $\text{F}$; (3) has (unconditional) preferences which are consistent with his conditional ones. These assumptions are shown to be sufficient to ensure an approximate representation of the decision maker's preference by a real-valued function W which has the form $\text{W(f) = v[}\text{Inf}{}_{\mathrm{\theta \in \because }}\text{E}{}_{\mathrm{\theta }}\text{(u°f),}\text{Sup}{}_{\mathrm{\theta \in \because }}\text{E}{}_{\mathrm{\theta }}\text{(u°f)]}$, where u and v, respectively, characterize the decision maker's attitudes toward risk and toward complete ignorance.     

Similar threshold functions from contrasting neural signal models

Three neural signal models of increment threshold detection are compared. All assume that the criterion for threshold is the attainment of a critical, minimum neural signal (or difference between two neural signals), and that the signal due to a test flash of intensity λ in the absence of a background light is $\text{λ}\text{(λ + σ)}$ (where σ is the semi-saturation constant). The models differ in the manner in which a background light of intensity θ is assumed to affect the signal. One model (due to Alpern et al., 1970a, Alpern et al., 1970b, Alpern et al., 1970c) assumes that the test flash signal, $\text{λ}\text{(λ + σ)}$, is attenuated by the multiplicative factor $\text{θ}{}_{\mathrm{D}}\text{(θ + θ}{}_{\mathrm{D}}\text{)}$ (where θ_D is a constant interpreted as sensory noise); another model specifies that the test flash signal is simply reduced (by subtraction) by the amount $\text{θ}\text{(θ + K)}$ (K a constant). One main result of this paper is that in the absence of pigment bleaching, these two models imply indistinguishable increment threshold functions. Further, a necessary and sufficient condition for each model guaranteeing the absence of saturation with steady backgrounds is found to be empirically satisfied. A third model is considered where the background field is assumed both to contribute to the neural signal and simultaneously to attenuate it (via a gain change). These assumptions are closely related to theoretical accounts of color induction and color perception. Though this model needs further investigation, it appears to be in better accord with actual increment threshold data than the others.     

An alternative scaling method for priorities in hierarchical structures

The early contributions of Saaty have spawned a multitude of applications of principal right (PR) eigenvector “scaling” of a dominance matrix [R]. Prior to Saaty's work (1977–1984) scaling of dominance matrices received little attention in multidimensional scaling, e.g., see Shepard (1972, pp. 26–27). This eigenvector method (EM) of scaling [R] yields u_i scores (weights) popularly used at each branching of the Analytic Hierarchy Process (AHP) technique that has been increasingly applied in multiple criterion analysis of utility, preference, probability, and performance. In this paper, it is proposed that an alternate least squares method (LSM) scaling technique yielding least squares optimal scores (weights) provides $\text{w}{}_{\mathrm{i}}{}^1$ values having a number of important advantages over u_i scores popularly utilized to date.     

Positive-difference structures and bilinear utility functions

Let (M₁, f), (M₂, g) be mixture sets and let ? be a binary preference relation on M₁ × M₂. By using the concept of positive-difference structures, necessary and sufficient conditions are given for the existence of a real-valued utility function u on M₁ × M₂ which represents ? and possesses the bilinearity property

$\text{u(?(α, x}{}_1\text{,x}{}_2\text{),g(β, y}{}_1\text{, y}{}_2\text{))}\text{=}\text{αu(x}{}_1\text{, g(βy}{}_1\text{, y}{}_2\text{))}\text{+}\text{(1 ? α) u(x}{}_2\text{, g(β, y}{}_1\text{, y}{}_2\text{))}\text{=}\text{βu(?(α,x}{}_1\text{, x}{}_2\text{),y}{}_1\text{)}\text{+}\text{(1 ? β) u(?(α,x}{}_1\text{, x}{}_2\text{),y}{}_2\text{)}$

, for all α, β ∈ [0, 1], all x₁, x₂ ∈ M₁ and all y₁, y₂ ∈ M₂. Moreover, uniqueness up to positive linear transformations can be proved for those utility functions. Finally an outline is given of applications of these results in expected utility theory.     

Difference measurement and simple scalability with restricted solvability

Let A, B be two sets, with B ? A × A, and ≤ a binary relation on B. The problem analyzed here is that of the existence of a mapping u: A → R, satisfying:

$\text{(a,b) ? (}\text{a}\text{?}\text{,}\text{b}\text{?}\text{)}\text{iff}\text{∨∧ μ(b) ? μ(a) ? μ(}\text{b}\text{?}\text{) ? μ(}\text{a}\text{?}\text{)}$

whenever (a, b), (a′, b′) ∈ B. In earlier discussions of this problem, it is usually assumed that B is connected on A. Here, we only assume that B satisfies a certain convexity property. The resulting system provides an appropriate axiomatization of Fechner's scaling procedures. The independence of axioms is discussed. A more general representation is also analyzed:

$\text{(a,b) ? (}\text{a}\text{?}\text{,}\text{b}\text{?}\text{)}\text{iff}\text{∨∧ F[μ(b), μ(a)] ? F[μ}\text{b}\text{?}\text{]}$

, where F is strictly increasing in the first argument, and strictly decreasing in the second. Sufficient conditions are presented, and a proof of the representation theorem is given.     

A generalization of comparative probability on finite sets

Let ? be a binary relation on a finite algebra $\text{A}$ of events A, B,…, where A ? B is interpreted as “A is more probable than B.” Conventional subjective probability is concerned with the existence of a probability measure P on $\text{A}$ that agrees with ? in the sense that A ? B ? P(A) > P(B). Because evidence suggests that some people's comparative probability judgments do not admit an agreeing probability measure, this paper explores a more flexible scheme for representing ? numerically. The new representation has A ? B ? p(A, B) > 0, where p is a monotonic and normalized skew-symmetric function on $\text{A}$ × $\text{A}$ that replaces P's additivity by a conditional additivity property. Conditional additivity says that $\text{p(A ? B, C) + p(?, C) = p(A, C) + p(B, C)}$ whenever A and B are disjoint. The paper examines consequences of this representation, presents examples of ? that it accommodates but which violate the conventional representation, formulates axioms for ? on $\text{A}$ that are necessary and sufficient for the representation, and discusses specializations in which p in separable in its arguments.     

On the scales of measurement

Let $\text{X}$ = 〈X, ≧, R₁, R₂…〉 be a relational structure, 〈$\text{X, ≧}$〉 be a Dedekind complete, totally ordered set, and n be a nonnegative integer. $\text{X}$ is said to satisfy n-point homogeneity if and only if for each x₁,…, x_n, y₁,…, y_n such that x₁ ? x₂ ? … ? x_n and y₁ ? y₂ … ? y_n, there exists an automorphism α of $\text{X}$ such that α(x₁) = y_i. $\text{X}$ is said to satisfy n-point uniqueness if and only if for all automorphisms β and γ of $\text{X}$, if β and γ agree at n distinct points of $\text{X}$, then β and γ are identical. It is shown that if $\text{X}$ satisfies n-point homogeneity and n-point uniqueness, then n ≦ 2, and for the case n = 1, $\text{X}$ is ratio scalable, and for the case n = 2, interval scalable. This result is very general and may in part provide an explanation of why so few scale types have arisen in science. The cases of 0-point homogeneity and infinite point homogeneity are also discussed.     

A representation theorem for finite random scale systems

This paper investigates necessary and sufficient conditions on choice probabilities P_a,B (of picking an element a in an offered set B), for the existence of random variables U_a, satisfying the equation P_a,B = $\text{P}${U_a = max {U_b | b ∈ B}} for all nonempty finite subsets B in a fixed set A, and all a ∈ B. A complete solution to this representation problem is obtained in the case where A is finite. The proof of the representation theorem provides an algorithm to construct the random variables U_a, up to some uniqueness properties. Investigation of these uniqueness properties show that an important part of the ordinal structure of the underlying random variables can be recovered.     

The psychology of inferring conditionals from disjunctions: A probabilistic study

There is a new probabilistic paradigm in the psychology of reasoning that is, in part, based on results showing that people judge the probability of the natural language conditional, if $A$then $B$, $P\left(ifAthenB\right)$, to be the conditional probability, $P(B\mid A)$. We apply this new approach to the study of a very common inference form in ordinary reasoning: inferring the conditional if not-$A$then $B$ from the disjunction $A$ or $B$. We show how this inference can be strong, with $P$(if not-$A$then $B$) “close to” $P\left(AorB\right)$, when $A$ or $B$ is non-constructively justified. When $A$ or $B$ is constructively justified, the inference can be very weak. We also define suitable measures of “closeness” and “constructivity”, by providing a probabilistic analysis of these notions.