Agreement vs. accuracy

Chapter 7 endnote 16, from How Emotions are Made: The Secret Life of the Brain by Lisa Feldman Barrett.
Some context is:

Suppose you and a friend are walking together when you see a man stamping his foot forcefully on the pavement. You categorize the man as angry. Your friend categorizes the man as dejected. [...] Who is correct in this case? [...] Emotions have no fingerprints, so there can be no accuracy. The best you can do is find consensus. We can ask other people if they agree with you or with me [...], or we can compare our categorizations to the norms of our culture.

It is probably most precise (although immeasurably more complicated) to say that the emoter (experiencing sadness) and the perceiver (seeing anger) are not literally conceptualizing the sensory input as the same event, because each has access to partially unique information (e.g.., his or her own physical state). They also might not share the same instances within their stored repository of prior knowledge about anger and sadness.

In a perceiver-dependent world, questions about “right” and “wrong” become transformed into questions about agreement. Social reality is rooted in social agreement.

• Do you and your friend agree in how you categorize the man's state? (Scientists call this inter-rater agreement.)
• Do you and the man agree in how you categorize him? (Scientists call this self-other agreement.)
• Do you both agree with your culture’s norms (usually represented by the experimenter's expectations)?

In addition to the ways of computing agreement that I list above, I could also use pattern classification (chapter 1) to build maps of neural activity for all three of you (you, your friend, the man), and see if the maps are classified the same. No matter how we measure, our questions are really about collective intentionality.

Notice that agreement replaces accuracy, because without an emotion fingerprint, accuracy can never be computed. The basic emotion method for studying emotion perception uses the language of “accuracy,” but as we explained in chapters 1 and 3, it really computes “agreement”; and so, it’s applying the logic of collective intentionality. Sometimes researchers compute agreement between the test subject viewing a posed facial configuration and the experimenter’s expectations. Sometimes it’s agreement between two subjects on what word they choose to categorize each with. Other times it’s agreement between a subject and a target person who’s being judged. In all these cases, “accuracy” is computed by consensus; and consensus is nothing but collective intentionality. In short, experimental methods like the basic emotion method are studying emotion as social reality, seemingly without realizing it.