Bridging Worlds

Notes after a conference in Castel Gandolfo on religion and development. If religious freedom and integral human development are graded concepts, the search for causes needs a logic of degrees.

For the past two days the International Mariapolis Centre in Castel Gandolfo hosted Bridging Worlds: Religion, Development, and the Future of Global Partnerships, a conference of the Atlantic Council’s Freedom and Prosperity Center, organised with the Center for Interreligious Dialogue of the Focolare Movement and with Sophia University Institute as partner and host, for which I served as conference secretary. It brought together scholars, development practitioners, policy makers and religious leaders from many traditions around a question that is becoming urgent.

The question

Faith-based institutions run a large share of the world’s schools, clinics and hospitals, hold land, manage savings and giving networks, and are often present where states are weak and international organisations struggle to stay. Most of this capacity operates outside the frameworks of mainstream development policy and finance. At the same time official development assistance is shrinking, several major donors are withdrawing, and confidence in top-down development models is low. The question of the conference was therefore how to build better partnerships between the development community and religious communities, and on what shared ground. The ground it proposed is integral human development, the idea from Catholic social teaching that development “cannot be limited to mere economic growth” and must “promote the good of every man and of the whole man”.1

James Mazzarella, who opened the conference for the Freedom and Prosperity Center, was candid about what is missing. We know that faith-based organisations matter, he said, but we struggle to show how much, in which settings, through which mechanisms and with what outcomes; nor should we assume that religious identity automatically confers effectiveness, legitimacy or accountability. His conclusion was simple: “We need better evidence. We need stronger analytical frameworks.” Fabio Petito, speaking for the academic committee, described the two lines of work that follow from this: mapping and measuring the faith-based development footprint, and listening to practitioners whose projects already embody a broader idea of development. He presented them as a form of mutual learning, in which religious actors learn the language of evidence and the policy world learns to attend to what it usually does not see.

In the spirit of that mutual learning, I would like to add a small philosophical note to the project, from the side of the philosophy of science. It is not a new way of measuring, but a reflection on the kind of concepts this research works with, and on the kind of logic that suits them.

Graded concepts

Religious freedom, integral human development and human flourishing are not concepts that a society either satisfies or fails to satisfy. They come in degrees: a country can have more or less religious freedom, a community can contribute more or less to development. They are also vague, in the technical sense that there is no sharp line at which a society stops being free in matters of religion. And they are multidimensional. The Pew Research Center, for instance, measures government restrictions on religion and social hostilities involving religion as two separate indices, because a country can score well on the first and badly on the second.2

None of this is a defect. These are simply the concepts we need in order to talk about what matters. However, they have two consequences for anyone who looks for causal arrows between them.

The first is that an index does not simply read off a concept that was already precise. It makes the concept precise, by choosing which dimensions count and how much each weighs. When the concept is vague, several choices are legitimate, and they need not agree. The divergence of ESG ratings, which often disagree about the same firm because they define and weigh sustainability differently, is a well-documented example.3 A causal relation found between two indices is therefore a relation between two precisifications of the concepts, and it holds for the concepts themselves only if it survives the reasonable alternatives.

The second consequence concerns the logic of the causal arrows themselves, and it deserves a closer look, even at the cost of a short detour.

A short detour into logic

Much of our talk about causes in the social world is talk about conditions. We say that religious freedom is necessary for flourishing, or that a strong civil society together with a functioning state is sufficient for it. To see what changes when the concepts are graded, it helps to start from the simplest logical tool we have, the arrow “if… then”, and to look at it slowly.

The classical arrow

In classical logic every statement is either true (1) or false (0). The arrow X→YX \rightarrow Y, read “if XX, then YY”, is defined by a small table:

XXYYX→YX \rightarrow Y
111
100
011
001

The arrow is false in one case only: when XX is true and YY is false. If XX is false, the arrow is true whatever the truth value of YY: it keeps a definite truth value, but places no constraint on YY.

This is all we need to talk about conditions. Saying that XX is sufficient for YY means that, in every case we look at, if XX then YY:

X→Y.X \rightarrow Y .

Saying that XX is necessary for YY means that there is no YY without XX, i.e. that in every case, if YY then XX. The arrow simply points the other way:

Y→X.Y \rightarrow X .

Every arrow can be read in two ways: in X→YX \rightarrow Y, XX is sufficient for YY, and YY is necessary for XX. Claiming that XX is sufficient for YY and claiming that it is necessary for YY are therefore two different claims, expressed by arrows that point in opposite directions: one can be true while the other is false. And in classical logic a single case in which the arrow fails, one country with XX and without YY, is enough to refute the claim that XX is sufficient for YY.

When truth comes in degrees

Now suppose that the statements themselves are graded. “This country is free in matters of religion” need not be simply true or false: it can be true to degree 0.8, or 0.3. Write xx for the degree of truth of XX and yy for that of YY, both between 0 and 1. Fuzzy logic, which grew out of Lotfi Zadeh’s work in the 1960s, tells us how to combine such degrees:4

Classical logicWith degrees of truth
not XXtrue when XX is false1−x1 - x
XX and ZZtrue when both are truethe smaller of xx and zz
XX or ZZtrue when at least one is truethe larger of xx and zz

When the degrees are only 0 and 1, these rules give back the classical ones exactly. The arrow can be generalised in the same spirit. A standard choice, due to the Polish logician Jan Łukasiewicz, is5

X→Y   is true to degree   min⁡(1,  1−x+y).X \rightarrow Y \;\text{ is true to degree }\; \min\big(1,\; 1 - x + y\big).

In words: the arrow is fully true as long as YY is at least as true as XX, and it loses truth as XX becomes truer than YY. Its table no longer has four rows, because xx and yy can take any value between 0 and 1, but a few rows show how it behaves:

xxyyX→YX \rightarrow Y
111as in classical logic
100as in classical logic
0.60.81YY is at least as true as XX
0.80.60.8a slight shortfall: the arrow is almost true
0.90.20.3a large shortfall: the arrow is mostly false

Three things that change

The first change is that the causal arrow itself becomes a matter of degree. A case in which XX is slightly truer than YY does not refute sufficiency; it weakens it. Charles Ragin, who turned these ideas into a method for comparative social research, summarises how well a group of cases supports the arrow X→YX \rightarrow Y with an index of consistency:6

consistency(X→Y)  =  ∑cmin⁡(xc, yc)∑cxc,\text{consistency}(X \rightarrow Y) \;=\; \frac{\sum_c \min(x_c,\, y_c)}{\sum_c x_c},

where the sums run over the cases cc. It equals 1 when, in every case, YY is at least as true as XX, and it falls as cases depart from this. The numbers matter less than what they show, so a toy example with four invented countries is enough:

Countryxx (religious freedom)yy (flourishing)smaller of the two
A0.90.80.8
B0.70.30.3
C0.40.50.4
D0.20.10.1
Sum2.21.71.6

The arrow “freedom → flourishing”, i.e. freedom is sufficient, has consistency 1.6/2.2≈0.731.6 / 2.2 \approx 0.73: country B has much freedom and little flourishing. The opposite arrow “flourishing → freedom”, i.e. there is no flourishing without freedom, has consistency 1.6/1.7≈0.941.6 / 1.7 \approx 0.94: only country C is slightly more flourishing than free. The same four countries support the second arrow far better than the first, and the logic makes the difference visible.

The second change concerns the principle that a statement and its negation cannot both be true. Classically, ”XX and not XX” is always false. With degrees, if “the country is free” is true to degree 0.6, then “the country is not free” is true to degree 0.4, and

X and not X   is true to degree   min⁡(0.6,  0.4)  =  0.4.X \text{ and not } X \;\text{ is true to degree }\; \min(0.6,\; 0.4) \;=\; 0.4 .

This is not a logical scandal. It is the formal counterpart of the plain observation that many societies are partly free and partly not, for instance free in law and hostile in practice, which is exactly what the two Pew indices record.

The third change is the most important for policy, and it is easiest to see with an example. Suppose that religious freedom is necessary for flourishing but far from sufficient: no country can flourish without being free, but being free does not guarantee flourishing. What would the data look like? Countries with little freedom would all have little flourishing. Countries with much freedom would be scattered: some would flourish a lot, others very little. On a graph with freedom on the horizontal axis and flourishing on the vertical one, all the countries would lie below the diagonal, in a triangle, as in the figure below.

Now ask two different questions of these data. The first is the one usually asked by statistical studies: when freedom increases, does flourishing increase too? The answer is “only weakly”, because among free countries flourishing varies a lot. The correlation is modest, and one might be tempted to conclude that freedom matters little. The second question is the one the arrow asks: is there any country that flourishes more than it is free? The answer is “none”. Yet the arrow “no flourishing without freedom” holds perfectly, and it is the one that matters most for anyone who wants to know what cannot be done without. The two questions are different because an arrow has a direction, while a correlation does not: a correlation measures only how two quantities vary together, and reads the same in both directions.

A necessary condition with a modest correlationIllustrative scatter plot of forty invented cases. The horizontal axis is the degree of a condition X, such as religious freedom; the vertical axis is the degree of an outcome Y, such as flourishing. All points lie on or below the diagonal Y equals X, so X is a perfectly consistent necessary condition for Y (consistency 1.00), while the Pearson correlation is only 0.30, because many cases with high X have low Y.000.50.511CONDITION X · e.g. degree of religious freedomOUTCOME Y · e.g. degree of flourishingY = XNECESSITYNo case lies abovethe diagonal: no case ismore Y than it is X.Consistency = 1.00CORRELATIONMany cases with high Xhave low Y.Pearson r = 0.30
An illustration with forty invented cases. The condition is perfectly necessary for the outcome, since no case lies above the diagonal, yet the correlation between the two is only 0.30.

Why this matters for religion

The difference is not only technical. Average effects assume, roughly, that a factor pushes in the same direction everywhere. A logic of graded conditions allows instead that the same condition contributes to an outcome only in combination with others, and that different combinations can lead to the same result. This is close to what Scott Appleby, one of the academic organisers of the conference, called the ambivalence of the sacred: the same religious resources can sustain violence or reconciliation, depending on the actors who interpret them and the circumstances in which they act.7 A logic of averages can easily lose sight of this ambivalence; a logic of degrees helps to keep it in view.

I do not want to overstate the point. These methods have their own difficulties, from the choice of the degrees assigned to each case to the limits of causal inference from a small number of cases, and vagueness can also be treated without fuzzy logic, for instance by asking what is true under every admissible precisification.8 The point is only that the choice of logic is part of the analytical framework, and that it is worth making it consciously.

Keeping our statements under control

What logic offers here is not a new body of evidence, but a way of keeping under control the statements we make. When we say that religion contributes to development, or that religious freedom matters for flourishing, it helps to know what exactly we are claiming. Is a condition sufficient, or necessary, or one ingredient among others? Is it present or absent, or present to a degree? Does the claim say that two things vary together, or that one cannot be had without the other? These are different statements, and they call for different kinds of evidence. Logic does not tell us which of them is true. It helps us see which one we are making, and what would count for or against it.

This seems to me a small contribution to the mutual learning Petito spoke of. Statements about religion and development travel between very different worlds, from religious communities to universities and policy institutions, and each world tends to hear them in its own way. Being clear about what kind of statement we are making is one way of making sure that, when they arrive, they still say what we meant.

There is a broader question here about what can and cannot be measured in this field at all, and about what the history of the natural sciences can teach it. I come back to it in the next note.

Cover: detail of the visual of the public roundtable “Religion, development, and peace: What is the future of global partnerships?”, Atlantic Council Freedom and Prosperity Center.

Footnotes

  1. Paul VI, Populorum Progressio (1967), §14. ↩

  2. Pew Research Center, Global Restrictions on Religion, with its Government Restrictions Index and Social Hostilities Index, published annually since 2009. ↩

  3. F. Berg, J. F. Kölbel and R. Rigobon, “Aggregate confusion: the divergence of ESG ratings”, Review of Finance 26 (2022), 1315–1344. ↩

  4. L. A. Zadeh, “Fuzzy sets”, Information and Control 8 (1965), 338–353. ↩

  5. On a technical note, fuzzy logic offers several ways of defining “and” and “if… then”, and the ones presented here are only one choice. Conjunction can be any of the operations logicians call t-norms: besides the minimum used in the table above (Gödel’s t-norm), common options are the product x⋅yx \cdot y and Łukasiewicz’s t-norm max⁡(0, x+y−1)\max(0,\, x + y - 1). Each t-norm comes with its own implication: the one used here belongs to Łukasiewicz’s t-norm, while the minimum goes with Gödel’s implication, which is 1 when x≤yx \le y and yy otherwise. All of them agree with classical logic when the degrees are 0 and 1, and all make the arrow fully true exactly when x≤yx \le y, which is what matters for the argument; the Łukasiewicz implication has the advantage of losing truth gradually. The choice is not innocent, though. With Łukasiewicz’s t-norm, for instance, ”XX and not XX” always has degree 0, so the second change described below holds for the minimum but not for every t-norm. See P. Hájek, Metamathematics of Fuzzy Logic, Kluwer, 1998. ↩

  6. C. C. Ragin, Fuzzy-Set Social Science, University of Chicago Press, 2000; Redesigning Social Inquiry: Fuzzy Sets and Beyond, University of Chicago Press, 2008. ↩

  7. R. S. Appleby, The Ambivalence of the Sacred: Religion, Violence, and Reconciliation, Rowman & Littlefield, 2000. ↩

  8. This is the supervaluationist approach; see K. Fine, “Vagueness, truth and logic”, Synthese 30 (1975), 265–300. ↩