Conceptual Analysis as Thickening:  Influence as an Example

1. Abstract

                 Conceptual Analysis as Thickening: Influence as an Example

Jingzhu Wei, School of Information Management, Sun Yat-sen University
Allen H Renear, School of Information Sciences, University of Illinois at Urbana-Champaign

Introduction

There has been a long conversation about method in the humanities and social sciences (Vico, Weber, Dilthey, Gadamer, and Winch, for example). Within the DH community this conversation takes a distinctive form amidst the interweaving of traditional humanistic approaches with the more empirical scientific methods associated with our new tools and research strategies[1,2,4,10]. Conceptual analysis can help support this interweaving by contributing to the forms available for the needed interpretive thickening of research in the digital humanities.

Thick description and conceptual analysis

Clifford Geertz makes a distinction, now classic in interpretative social science, between “thin” and “thick” description[6], terms first used by Gilbert Ryle, who offers this example: a thin description would describe a composer as producing note sequences, but a thick description interprets the same phenomena as “cancellings, modifyings, assemblings, reassemblings, rehearsings” — that is, the composer’s thoughts and intentions situated within a particular cultural context[12].

Conceptual analysis clarifies a concept by formally identifying conditions individually necessary and collectively sufficient for its occurrence. Although central to analytic philosophy, in the last several decades conceptual analysis has also been used to complement empirical methods in the social and cultural sciences, and, more recently, in the information sciences[3,5,8,9,14,16].

For cultural phenomena the requirements identified by conceptual analysis can help provide opportunities for thickening descriptions. We illustrate this with the concept of (intellectual) influence.

Influence in the humanities, and in the digital humanities

Much research in the humanities explores influence — how the literary, religious, or scientific beliefs, attitudes, tastes, or feelings of a person or group of persons influenced others, often focusing on artifacts or events considered as evidence for influence, instruments of influence, or the primary agents and patients.

In the digital humanities our projects explore influence by putting data in digital form and analyzing that data with computational methods[15].   We detect stylistic similarities, transmitted corrections and errors, related themes and topics, and so on. We then advance claims or hypotheses that assert or explain the influences we have evidence for.

Although understanding influence is a common research objective there has been little effort to define the concept itself. We routinely indicate what we count as evidence of influence, but rarely say, clearly and exactly, what we mean by the term — such omissions will become increasingly problematic in the future as research in the digital humanities, and in fact the humanities in general, is likely to be focusing on larger and larger quantities of thinner and thinner data.  

Motivation for conceptual clarification: Causation is not Influence

We are often told that correlation is not causation — but neither is causation influence, and thinning data and methods are conducive to the latter fallacy as well. We might say a person X influenced a person Y because X’s views had a causal effect on Y’s views. Suppose however that a novelist’s views on class lead to a successful potboiler and the proceeds are anonymously donated to a political theorist — who then has time to develop a particular analysis of class. This is causation to be sure, and a quick inference from a thin description might classify it as influence, but it is not influence — influence is causation of a certain sort.

The example may seem cooked up, but it makes the point. The well-known problem of “deviant causal chains”[11] is a challenge both to empirical methods, which must extract influence from causation, and to analytical efforts which attempt to identify a basis for that distinction. These are related issues. Preventing the conflation of causation and influence in a world of thin data, requires a deeper conceptual understanding of influence.

Steps towards a conceptual analysis of influence

We first adapt Grice’s account of intended meaning[7]:

S states that P if and only if

    S utters U intending that

1. Someone, x, forms the belief that P
2. x recognizes that S intends 1)
3. x forms the belief that P at least partially because 2)

Grice establishes that neither 1) alone, nor 1) and 2) together are sufficient for an occurrence of some particular linguistic behavior meaning P, and although some problems remain the three clauses together to appear to be a reasonable first characterization. We therefore begin our account of influence by building on Grice:

H is influenced by S =df

1. S states that P
2. H forms the belief that P
   and does so at least partially because H recognizes that S stated that P

Such a limited notion of influence may seem of little interest and we make no claims for direct applicability. A complete and robust account of influence would consider influences from/on sensibility, taste, and other affective a states, as well as capabilities, skills, propagated or transitive influence, and more; it would also include the influence of communicative objects and non-linguistic events.

Nevertheless, these analyses establishes much: the need for intention, the need for reflexive intention (intentions about intentions), and, in particular, for an intention that recognition of intention be a partial cause of the response. Such features can help guide the construction of thick interpretive description from thin digital data.

Deviant Causal Chains: Resolved and Restored

Another achievement of this analysis is the elimination of a class of deviant causal chain counterexamples. In the case described above the political theorist’s views will fail our requirement that they be at least partially caused by the recognition of the novelist’s views, and so will not be counted as influence.

Unfortunately another subset of cause/influence conflating counterexamples remains: suppose the novelist communicates his views on class to the political theorist, who then herself writes and publishes the potboiler (based on those views) that funds her research. The recognition condition is met, but the theorist’s research has not been (intellectually) influenced by the views of the novelist.

Interpretation Abides

Eliminating deviant causal chains counterexamples has proven difficult in other analyses of fundamental social and cultural concepts, and may be impossible. But this does not undermine the practical usefulness of conceptual analysis — rather it reminds us of the limitations of formal strategies, and the fluid nature of conceptual understanding. It is significant that we rarely have any trouble with cases: once we understand the case we recognize immediately whether the causal chain is deviant or legitimizing. Conceptual analysis can help guide thick description in the digital humanities, but it is not a replacement for interpretation.

References

1. Ciula, Arianna, Cristina Marras, Circling Around Texts and Language: Towards “Pragmatic Modelling” in Digital Humanities. Digital Humanities Quarterly 10:3, 2016.
2. Clement, Tanya E., Where Is Methodology in Digital Humanities? In: Debates in the Digital Humanities, University of Minnesota, ed. Matthew K. Gold and Lauren F. Klein, 2016.
3. Derr, Richard L., A Conceptual Analysis of Information Need, Information Processing & Management 19:5 1983.
4. Domenico, Fiormonte, Valentina Martiradonna, Desmond Schmidt. Digital Encoding as a Hermeneutic and Semiotic Act: The Case of Valerio Magrelli. Digital Humanities Quarterly 4:1, 2010.   
5. Furner, Jonathan, Conceptual Analysis: A Method for Understanding Information as Evidence, and Evidence as Information. Archival science 4:3-4, 2004.
6. Geertz, Clifford, Thick Description: Toward an Interpretive Theory of Culture. In: The Interpretation of Cultures, Basic Books, 1973.
7. Grice, H. P., Meaning. The Philosophical Review 66:3 1957.
8. Korman, Dan. Z., Eric Mack, Jacob Jett, Allen Renear, Defining Textual Entailment. Journal of the Association for Information Science and Technology 69, 2018.
9. Lee, Jin Ha, Allen H Renear, and Linda C Smith, Known?Item Search: Variations on a Concept. Proceedings of the Association for Information Science and Technology, 43, 2006.
10. Liu, Alan, Where Is Cultural Criticism in the Digital Humanities? In: Debates in the Digital Humanities University of Minnesota, ed. Matthew K. Gold, 2013.
11. Peacocke, Christopher, Deviant Causal Chains. Midwest Studies in Philosophy 4, 1979.
12. Ryle, Gilbert, Thinking and Reflecting. Royal Institute of Philosophy Supplements 1968.
13. Shope, Robert. The Analysis of Knowing: A Decade of Research. Princeton, 1983.
14. Savolainen, Reijo, Emotions as Motivators for Information Seeking: A Conceptual Analysis. Library and Information Science Research 36:1, 2014.
15. Wei, Jingzhu, Rui Liu, An Approach to Constructing a Knowledge Graph of the Hundred Schools of Thought in Ancient China. ACM/IEEE Joint Conference on Digital Libraries (JCDL), 2019.
16. Wickett, Karen. M., Allen H. Renear, Jonathan Furner, Are Collections Sets? Proceedings of the Association for Information Science and Technology 48: 2011.

Relation to conference theme: Thick description is recognized as supporting cultural understanding, especially for observers outside the studied culture; conceptual analysis adds a format that can be used for explicitly and iteratively negotiate difference and agreement.