Aaron
@aaronwang_23
Followers
159
Following
50
Media
6
Statuses
38
MD Candidate @PennMedicine | @JohnsHopkins alum
Joined February 2014
Think of the last time you learned something new. Did you learn from someone else? Could you have learned it on your own? In this preprint, we ask how humans learn in different environmentsâfrom getting full instructions to almost no instructions at all!
osf.io
Humans often learn about the world by observing, imitating, and receiving explicit instruction from others. Even when learning via trial-and-error or reflection, human adults often do so immersed in...
2
26
59
1/ A century ago, Tolman showed that meaningful learning can occur even without apparent performance improvements. What is the neural basis of this distinction between learning something vs learning to perform? We tackle this in a new manuscript đ„ đđŹđ° https://t.co/DXQHjNRSpR
5
41
156
@PennNSG Kudos to Aaron Wang @aaronwang_23 who completed his capstone project on #ArtificialInteligence in Acute #Neurosurgical Pathway - Practical Imple...
0
1
5
Have you ever wondered when a habit becomes a habit? Like when you were supposed to turn left to the grocery store but ended up turning right as if going to work? Check out our recent work w/@kishoreneuro @baselesspursuit @ziyizhu15 @Sun_Ruolan and twitter-less Zyan Wang in a đ§”
2
23
68
In an exciting new direction for the lab, we ( https://t.co/zeb6oen1oH) teamed up with Cindy Moss ( https://t.co/VaTGafBlNb) and Mel Wohlgemuth ( https://t.co/5F8parEN06) to bring two-photon imaging to the echolocating bat! Excited to share our preprint about this work (1/13)
Spatially clustered neurons encode vocalization categories in the bat midbrain https://t.co/CGxQ2msG2r
#biorxiv_neursci
4
36
119
So interesting!
Think of the last time you learned something new. Did you learn from someone else? Could you have learned it on your own? In this preprint, we ask how humans learn in different environmentsâfrom getting full instructions to almost no instructions at all!
0
2
5
Cool new work from @aaronwang_23 @tandom_lauren Damola Ongunlade @jorgemlg and @kishoreneuro! Had the pleasure of seeing this project unfold up close, excited to see it out in the wild
Think of the last time you learned something new. Did you learn from someone else? Could you have learned it on your own? In this preprint, we ask how humans learn in different environmentsâfrom getting full instructions to almost no instructions at all!
0
3
11
Excited to share this new work from the lab! Our first foray into human learning experiments. Read @aaronwang_23 tweeprint to learn more about how humans learn without instructions in a simple task (it may surprise you!) Just a few words about the team that made this happen (1/5)
Think of the last time you learned something new. Did you learn from someone else? Could you have learned it on your own? In this preprint, we ask how humans learn in different environmentsâfrom getting full instructions to almost no instructions at all!
1
13
51
All the data, code, analysis, and pre-registrations can be found here: https://t.co/oWkOag80Al. This project was co-lead with @tandom_lauren, and in collaboration with Damola Ogunlade, @jorgemlg and @kishoreneuro.
0
0
3
Future work should explore how these variables interact in more complex, real-world scenarios where historically disadvantaged backgrounds, ambiguous feedback, and/or hidden curricula are at play.
1
0
1
Here, actionable instructions that reduce exploration may produce more efficient learning outcomes compared to laying out a more general, albeit abstract and therefore not immediately actionable picture of the task at hand.
1
0
1
It is easy to underestimate the crucial role played by subtle background knowledge, explicit instruction, and luck in learning even the simplest of tasks.
1
0
1
Overall, our data show that small changes in the nature and content of instructions (in our case, explicit limits on action exploration) can lead to profound changes in final performance and cross-subject variability.
1
0
1
In the last three experiments, we replicated the same pattern of results with an object recognition task. The tasks were identical except that instead of withholding pressing a key, there was a second key ("enter") that subjects need to map to a specific pair of shapes.
1
0
1
Maybe there is something unique to âwithholdingâ that is difficult for people to learn through instructions? Even when subjects were told that not pressing any key was one valid option, people continued to explore different key presses during the task.
1
0
2
Surprisingly, telling subjects about the abstract go/no-go structure of the task did not produce noticeable learning effects. In other words, knowing that in some trials they would need to press some key and in others they would need to withhold didn't help subjects in the task.
1
0
1
Only subjects who received explicit information identifying the spacebar as the relevant key showed improved performance compared to zero instructions. Implicit spacebar training increased subjects' bias to press the spacebar more often, but it didn't help them learn better.
1
0
1
Subjects were provided with one of the following: behavioral shaping of the relevant action (e.g., spacebar training), explicit (written) information identifying the spacebar as the relevant key, or explicit information only detailing the go/no-go structure of the task.
1
0
1
To explore how varying degrees of instruction detail may improve behavioral performance compared to receiving zero instructions, we ran 3 additional experiments using the same go/no-go paradigm.
1
0
1
Importantly, luck played a crucial role! Getting that first spacebar press or withhold correct predicted a higher performance on that type of trial ("go" for spacebar, "no-go" for withhold). But whether that first action is correct or not depends on chance!
2
0
2
In contrast with the full instructions experiment, more than half of the subjects failed to reach expert performance when they didn't receive instructions despite the extreme simplicity of the task itself.
1
0
3