Jev almost never calls anyone furious or terrified

When a feeling comes in a strong and a mild word (furious or angry, terrified or afraid, devastated or sad), which one does Jev use?

result2,982 questions

Jev almost never calls anyone furious, and seldom terrified. Of 90 stories people wrote about feeling furious, Jev calls 3% furious and 56% angry; of 86 about feeling terrified, 19% terrified and 63% afraid. It rarely goes the other way: 1% of angry stories become furious, 10% of afraid ones terrified. Devastated and sad are mixed up equally both ways (18% each).

furious / angry1% · 56%

terrified / afraid10% · 63%

devastated / sad18% · 18%

mild word read as strongstrong word read as mild

How to read this: One pair of bars per feeling. Grey: stories written for the strong word that Jev called by the mild one. Magenta: the reverse, stories written for the mild word that it called by the strong one. The gap between them is how much Jev turns the volume down.

557 stories in the three pairs; 90% interval on furious read as angry [0.478, 0.644]. Across all 32 labels, the words Jev uses least relative to writers are furious (4 uses for 90 stories), ashamed (9 uses for 94 stories), sentimental (22 uses for 95 stories), faithful (25 uses for 95 stories), terrified (33 uses for 86 stories).

In short

Jev hears strong feelings as milder ones: stories about being furious mostly come out "angry", stories about being terrified mostly "afraid".
The softening runs one way. Mild stories are rarely turned up to the strong word.
Furious is the word Jev uses least next to how often writers were given it: 4 times for 90 stories.

What the data shows

Sleeping Shaq meme: a story about feeling furious; Jev: angry, I guess

Furious becomes angry. 56% of the furious stories are called angry; only 3% are called furious. The reverse happens to 1% of angry stories.
Terrified becomes afraid. 63% of terrified stories are called afraid, 19% terrified; 10% of afraid stories go the other way.
Devastated and sad are confused equally both ways: 18% each, no direction.
Furious is Jev's rarest word: it used it 4 times for 90 furious stories. Ashamed (9 uses for 94 stories) is next.

What it means, and what it doesn't

When Jev names a feeling, it reaches for the moderate word. In summaries and replies, people's strongest emotions would come out toned down, a consistent softening rather than random error. The one pair that doesn't soften, devastated and sad, suggests it's about the words themselves: "furious" and "terrified" are the words it avoids.

It doesn't mean Jev misses that the writer is angry or afraid; it mostly gets the family right. It misses how strongly.

Caveats

A story written to a word. Each writer was handed an emotion word and asked to describe a time they felt it. A story written for "furious" may genuinely read as plain anger; the label is the prompt, not a measurement of intensity.
No other readers to compare. Jev is compared with the word the writer was given, not with how other people would label the same story. Other readers might soften these stories too.
Many near-synonyms. Jev chose from 32 words, including angry, annoyed and furious. With that many close options, any reader will spread its answers; what matters is that Jev's errors all go one direction.

Jev on this experiment

Would a person find it interesting to read?: Yes76%
Does it describe you?: No54%
Would you have predicted it?: No59%
How fair is the comparison?: The comparison is reasonable
How much should a reader rely on it?: Moderately
Which caveat matters most?: A story written to a word52%

Why ask this

Feelings come in strengths. Furious is more than angry, terrified is more than afraid, devastated is more than sad. When a reader picks the milder word, it quietly turns the volume down on what the person said.

That matters for any model that summarizes feedback, triages messages or writes back to people: if it hears fury as annoyance and terror as worry, its summaries will read calmer than the people who wrote them.

How this was done

The people and the data

EmpatheticDialogues (Rashkin and colleagues, 2019), a dataset from Facebook AI researchers: crowd workers were each given one of 32 emotion words and wrote a situation from their own life in which they felt it, in one to three sentences. Three of the 32 come in a strong and a mild version of the same feeling: furious and angry, terrified and afraid, devastated and sad. That gives 557 stories across the three pairs, about 90 per word.

What Jev was asked

The situation, and all 32 words to choose from:

Someone wrote [situation] about a time in their own life. Which emotion were they feeling?
sad · angry · proud · afraid · caring · guilty · joyful · ... · furious · ... · terrified · devastated · ... (32 words in all)

For example, someone wrote under "furious": "I won tickets to a concert and when we got there, they were supposed to have the tickets at the box office and they didn't so the refused us no matter how much proof we gave that we won. It was ridiculous!" Jev said angry (82%).

How it was measured

For each pair, how many stories written for the strong word Jev calls by the mild one, and how many written for the mild word it calls by the strong one. Across all 32 words, the analysis also compares how often Jev uses each word with how often writers were given it.

Where these questions live

2,982 questions across 1 topic of the map. Each opens on the map with every question in it.

Reading Feelings & Character

Emotions in Personal Stories2,982

Every question

All 2,982 questions behind this result, the telling ones first: the examples the analysis points to, then the ones where Jev misses, biggest gap first.

Jev’s own answer

Showing 0 of 2,982