2026-01-26¶
- Date: 2026-01-26
- Lead discussion: Dania Machlab
- Paper: Lerner, Talia N., Ashley L. Holloway, and Jillian L. Seiler. "Dopamine, updated: reward prediction error and beyond." Current opinion in neurobiology 67 (2021): 123-130 Article
Notes¶
Questions¶
- Q: What grade on a scale from 1 (worst) to 10 (best) would you give this paper?
My answer
8
- Q: How would you praise the paper?
My answer
It feels like a friendly, honest, open, impartial review.
- Q: How would you criticise the paper?
My answer
Not all of the references have a mini-summary. If I would be suspicious/paranoid, I would investigate if there is some partiality there.
- Q: How would you summarize this paper in one line?
My answer
How neurons behave and learn is a dynamic field
- Q: Should we do what is in the paper?
My answer
Enjoy the enthusiasm of its authors.
- Q: How does this paper make us a better teacher?
My answer
I find it hard to use these concepts to teaching.
I do like the A -> R, after which AX -> R has no effect:
I feel that I too wold ignore the additional/useless signal.
I am unsure if we say something like this: When we first teach a way that works, it is useless to present another way that works without extra information. Else, on a neuronal level, no learning happens.
Scribbles¶
| Abbreviation | Full |
|---|---|
| RPE | Reward prediction error |
| VTA | Ventral tegmental area |
I enjoy that the TLC tries to find papers that give good evidence in becoming a better teacher. At the start of reading this paper, I wonder how this will help us become a better teacher.
I like how they add comments to the references!
Study 4¶
| Procedure | Results |
|---|---|
1. Train A -> R |
VTA dopamine neurons give RPE |
2a. Train AX -> R |
'Blocking': X does not change the behavior of the neuron |
2b. Train AX -> R and stimulate neurons to give RPE |
A -> R is unlearnt. Conclude: RPE signal updates learning system |
In my words: if a neuron has an association, additional info is ignored, unless the neurons are artificially stimulated to 'chemically pretend' to be learning.
Study 5¶
| Procedure | Results |
|---|---|
1. Train A -> R |
VTA dopamine neurons give RPE |
2. Train AX -> R |
'Blocking': X does not change the behavior of the neuron |
3. Train AY -> R |
'Blocking': Y does not change the behavior of the neuron |
4. inhibit neurons during AX (but not AY) |
Learning about X and Y was equally blocked. |
In my words: if a neuron has an association, additional info is ignored. This additional info does not give any extra chemical signal whatsoever.
From 10: 'if dopamine release is inhibited during a novel cue, learning about that cue is impaired'
'RPE signals are not necessarily accurate representations of error, but reflect an animal’s internal understanding of its environment based on previous experiences in that environment'
Ramping: I would explain this is as: When an interesting target becomes available, a high RPE is given off, which decreases when the subject is getting closes. The decrease in RPE is the chemical representation of 'motivation'.
Ramps are 'a reduction in task dimensionality as subjects narrow their focus to relevant stimuli'.
Ramps exist 'where internal goal representations are necessary to support behavior'.