Skip to contents

classroom

classroom.Rmd

Teaching with a public-domain corpus

Aesop’s fables are short, narrative, and structured around morals—which makes them unusually good for research methods teaching:

  • Operationalization: “What counts as sentiment?” “What counts as a theme?”

  • Measurement decisions: Tokenization, stop words, lexicon choice

  • Transparency & replication: Same corpus, different pipelines, different conclusions

  • Critical thinking: What do methods miss in moral storytelling?

We’ll use one fable as a running example: “The Fox and the Grapes.”

library(aesopR)

Helper: pick the fable (“Fox and the Grapes”) 

fox_rows <- grepl("fox", aesops_fables$title, ignore.case = TRUE) &
  grepl("grape", aesops_fables$title, ignore.case = TRUE)

fox_grapes <- aesops_fables[fox_rows, ]
fox_grapes[, c("fable_id", "title", "moral")]
fable_id <- fox_grapes$fable_id[1]
fable_id

Activity 1: “Guess the moral” from partial evidence

Goal: Show how conclusions depend on what evidence you allow yourself to see.

Setup (5 minutes):

  1. Give students only the title and the top 10–15 words.

  2. Ask them to write a one-sentence “predicted moral.”

  3. Then reveal the full moral and discuss mismatches.

Instructor code: top words 

fox_tokens <- aesops_tokens |> filter(fable_id == "005")
top_words <- fox_tokens |> count(word,sort = TRUE) |> slice_head(n = 10)
top_words

Prompt (students):

  • Based on the title and these words, what do you think happens?

  • What do you think the moral is?

  • What words might be misleading?

Discussion points:

  • Frequency ≠ importance

  • Stop words and narrative function words

  • “Theme inference” is not purely mechanical

Activity 2: “Methods as measurement” (lexicon comparison)

Goal: Demonstrate that measurement depends on instrument choice (lexicon choice).

Quick prompt (students):

  • Do these methods agree about whether the fable is “positive” or “negative”?

  • What kinds of words do they treat as meaningful?

Bing sentiment counts 

fox_bing <- aesops_bing |> filter(fable_id == "005")
table(fox_bing$sentiment)

AFINN score summary 

fox_afinn <- aesops_afinn |> filter(fable_id == "005")
summary(fox_afinn$value)

Identify sentiment-driving words (simple) 

# Bing: which words appear most in each label?
sort(table(fox_bing$word[fox_bing$sentiment == "negative"]), decreasing = TRUE)[1:10]
sort(table(fox_bing$word[fox_bing$sentiment == "positive"]), decreasing = TRUE)[1:10]
# AFINN: strongest positive/negative words
afinn_ordered <- fox_afinn[order(fox_afinn$value), c("word", "value")]
head(afinn_ordered, 8)
tail(afinn_ordered, 8)

Discussion points:

  • Lexicons were built for different purposes

  • Context blindness (sarcasm, negation, narrative framing)

  • Reliability/validity analogy: lexicon as “instrument”

Activity 3: “Operational definitions” (what counts as a construct?)

Goal: Have students define a construct (e.g., “frustration” or “self-justification”)

and propose a text-based operationalization.

Small-group task (10–15 minutes):

Pick one construct and propose:

  1. An operational definition in plain language

  2. A rule-based measurement plan using tokens/sentiment

  3. A limitation statement

Example constructs:

  • Frustration

  • Perseverance

  • Envy

  • Self-justification (“sour grapes”)

  • Rationalization

Instructor code: create simple “dictionary” counts

Below is a simple, transparent “dictionary method” that students can critique.

# Tiny hand-built dictionary (edit in-class)
rationalization_terms <- c("couldn't", "cannot", "never", "worth", "sour", "fine")

# Count dictionary hits
sum(fox_tokens$word %in% tolower(rationalization_terms))

Prompt (students):

  • Are these terms reasonable indicators?

  • What false positives/negatives might occur?

  • How would you improve the dictionary?

Discussion points:

  • Face validity vs construct validity

  • Researcher degrees of freedom

  • Pre-registration and transparency

devtools::document()
devtools::check()
pkgdown::build_site()