Interpretable abstractions of artificial neural networks predict behavior and neural activity during human information gathering.
15 blocks of code here have a permanent identifier (oscr:b:…), listed with the matches below: citable for good, with a page of its own. Select a match and press Cite to copy it. What a block identifier is.
The 17 matches · 2 of them tie a paragraph to a whole file without given lines: weak matches, whose lines are not tinted
- [1] § Methods › Hybrid model ↔ HybridModelling/src/analysis/recover/recover.jl, lines 139 to 199 · score 0.72 · neural network, cognitive model, hybrid model, Adam, cm, nn · block
oscr:b:VXEBRQYSA4QS4-E· Report a mismatch - [2] § Methods › Hybrid model ↔ HybridModelling/src/analysis/fit/hybrid-model/fit-hybrid.jl, lines 54 to 127 · score 0.67 · cognitive model, hybrid model, Adam, cm, nn, minimizing · block
oscr:b:PVANF0SV64WZB-F· Report a mismatch - [3] § Results › Sampling behavior adaptively scales with task difficulty and uncertainty ↔ BehavioralModelling/src/speed-accuracy/trial-level/green-dots.jl, the whole file · a weak match · score 0.66 · Poisson, mixed, blocked, behavior, unattended, visit · block
oscr:b:6PDM7HA71AD5Z-9· Report a mismatch - [4] § Results › The ANN can be transformed into an interpretable symbolic function ↔ HybridModelling/src/analysis/symbolic-regression/cross-task-generalization/compare-loss.jl, lines 64 to 137 · score 0.61 · Wilcoxon signed rank, Cross task generalization, symbolic model, symbolic regression, UCB model, median · block
oscr:b:CH5D2933XB9CW-6· Report a mismatch - [5] § Methods › Markov decision process–based model ↔ OptimalPolicy/src/mdp.jl, lines 132 to 197 · score 0.60 · beta binomial, reward function, transitions, policy, optimal · block
oscr:b:P8WGDX2F9TS19-O· Report a mismatch - [6] § Results › The ANN-derived VoI predicts participants’ sampling decisions ↔ HybridModelling/src/analysis/compare/losses/compare-losses.jl, the whole file · a weak match · score 0.60 · cross validation, loss relative, linear model, symbolic model, UCB model, trained · block
oscr:b:G22VMDSVFASCX-6· Report a mismatch - [7] § Methods › fMRI data preprocessing ↔ BrainModelling/first-level/script/feat/create_design.py, lines 13 to 144 · score 0.59 · high pass filtered, motion, template, Preprocessing, temporally, Brain · block
oscr:b:RJN87H1686QBQ-Q· Report a mismatch - [8] § Results › The ANN can be transformed into an interpretable symbolic function ↔ HybridModelling/src/analysis/symbolic-regression/cross-task-generalization/compare-loss.jl, lines 245 to 300 · score 0.58 · Wilcoxon signed rank, symbolic regression, median, UCB, ANN · block
oscr:b:HEZ0WBWFHK284-T· Report a mismatch - [9] § Methods › Symbolic regression ↔ HybridModelling/src/analysis/symbolic-regression/after-first-visit/estimate.jl, lines 112 to 135 · score 0.56 · combined score, Symbolic regression, loss, visit, switching, hybrid · block
oscr:b:WGADBXEJ18N7M-B· Report a mismatch - [10] § Methods › fMRI data analysis ↔ BrainModelling/first-level/script/feat/create_design.py, lines 13 to 144 · score 0.55 · motion parameters, confound, volumes, temporal, preprocessing, voxel · block
oscr:b:RJN87H1686QBQ-Q· Report a mismatch - [11] § Methods › Symbolic regression ↔ HybridModelling/src/analysis/symbolic-regression/interpret/make-figure.jl, lines 1 to 51 · score 0.55 · Symbolic regression, switch function, Na, Nu, jl, visit · block
oscr:b:N2KMN6TEQRGH3-T· Report a mismatch - [12] § Results › Sampling behavior adaptively scales with task difficulty and uncertainty ↔ OptimalPolicy/src/solve.jl, lines 1 to 43 · score 0.54 · optimal policy, dynamic programming · block
oscr:b:EQVG7JDEZH905-M· Report a mismatch - [13] § Methods › Markov decision process–based model ↔ OptimalPolicy/src/solve.jl, lines 1 to 43 · score 0.54 · backward induction, recursion, policy, optimal · block
oscr:b:EQVG7JDEZH905-M· Report a mismatch - [14] § Methods › Behavioral analysis ↔ HybridModelling/src/analysis/fit/symbolic-model/symbolic2ann/make-figure.jl, lines 353 to 411 · score 0.53 · voiU, voiA, logistic, fitted, unattended, ANN · block
oscr:b:H4R8ZRSWPBD1V-E· Report a mismatch - [15] § Methods › ROI analysis ↔ BrainModelling/second-level/script/post-feat/roi/fit-data/fit-model.jl, lines 1 to 53 · score 0.53 · effsize, ROI, cope, voxel, mixed, formula · block
oscr:b:6PRJMP5M60WTV-Q· Report a mismatch - [16] § Results › The ANN-derived VoI predicts participants’ sampling decisions ↔ HybridModelling/src/analysis/fit/cross-validation.jl, lines 1 to 31 · score 0.52 · cross validation, linear model, symbolic model, UCB model, loss, fit · block
oscr:b:RPSQ280Y3204R-5· Report a mismatch - [17] § Methods › Behavioral analysis ↔ HybridModelling/src/analysis/fit/symbolic-model/symbolic2ann/fit.jl, lines 96 to 165 · score 0.51 · voiU, voiA, fitted, switching, unattended, visit · block
oscr:b:DR64YKMRCZCF0-T· Report a mismatch
Paper
Loaded from Europe PMC by your browser; OSCR does not store it: doi.org · Europe PMC
The paper is loaded when this pane is shown.
The authors' code
Julia · 300 lines · 9.1 KB · MIT · 2 matches
- using YAML, CSV, DataFrames, Statistics, GLMakie, Colors, HypothesisTests, CairoMakie;
- config = YAML.load_file("config.yaml");
- include(joinpath(config["mac"]["project"], "helper.jl"));
- # Dataset 1 ------------------------------------------------------------
- loss_symbolic = [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "Symbolic-model", "loo", "outcome",
- "gershman2018-data1", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:45
- ] |> x -> deleteat!(x, 18)
- loss_ucb = [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "ucb-model", "loo", "outcome",
- "gershman2018-data1", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:45
- ] |> x -> deleteat!(x, 18)
- loss_aucb = [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "aucb-model", "loo", "outcome",
- "gershman2018-data1", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:45
- ] |> x -> deleteat!(x, 18)
- begin
- fig = Figure(size = (600, 400));
- ax = GLMakie.Axis(fig[1, 1], ylabel = "Loss", xlabel = "Subject", limits = (0, 45, 0, 1));
- lines!(ax, loss_symbolic, label = "Symbolic");
- lines!(ax, loss_ucb, label = "UCB");
- lines!(ax, loss_aucb, label = "AUCB");
- axislegend(ax; position = :lt);
- fig
- end
- begin
- fig = Figure(size = (600, 400));
- ax = GLMakie.Axis(fig[1, 1], ylabel = "Loss", xlabel = "Subject");
- density!(ax, loss_symbolic .- loss_ucb)
- fig
- end
- begin
- fig = Figure(size = (600, 400));
- ax = GLMakie.Axis(fig[1, 1], ylabel = "Loss", xlabel = "Subject");
- density!(ax, loss_symbolic .- loss_aucb)
- fig
- end
- # Sign-rank test Symbolic vs UCB ------------------------------------------------------------
- signedrank_test_data1 = SignedRankTest(loss_symbolic, loss_ucb);
- print_summary(signedrank_test_data1);
- """
- Wilcoxon signed-rank test
- W = 4.0, n = 44, median = -0.07, P = 7.96 × 10⁻¹³
- """
- # Sign-rank test Symbolic vs aUCB ------------------------------------------------------------
- signedrank_test_data1 = SignedRankTest(loss_symbolic, loss_aucb);
- print_summary(signedrank_test_data1);
- """
- Wilcoxon signed-rank test
- W = 163.0, n = 44, median = -0.01, P = 4.81 × 10⁻⁵
- """
- # Dataset 2 ------------------------------------------------------------
- loss_symbolic = [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "Symbolic-model", "loo", "outcome",
- "gershman2018-data2", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:44
- ]
- loss_ucb = [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "ucb-model", "loo", "outcome",
- "gershman2018-data2", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:44
- ]
- loss_aucb = [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "aucb-model", "loo", "outcome",
- "gershman2018-data2", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:44
- ]
- begin
- fig = Figure(size = (600, 400));
- ax = GLMakie.Axis(fig[1, 1], ylabel = "Loss", xlabel = "Subject", limits = (0, 44, 0, 1));
- lines!(ax, loss_symbolic, label = "Symbolic");
- lines!(ax, loss_aucb, label = "AUCB");
- lines!(ax, loss_ucb, label = "UCB");
- axislegend(ax; position = :lt);
- fig
- end
- begin
- fig = Figure(size = (600, 400));
- ax = GLMakie.Axis(fig[1, 1], ylabel = "Loss", xlabel = "Subject");
- density!(ax, loss_symbolic .- loss_ucb)
- fig
- end
- begin
- fig = Figure(size = (600, 400));
- ax = GLMakie.Axis(fig[1, 1], ylabel = "Loss", xlabel = "Subject");
- density!(ax, loss_symbolic .- loss_aucb)
- fig
- end
- # Sign-rank test Symbolic vs UCB ------------------------------------------------------------
- signedrank_test_data2 = SignedRankTest(loss_symbolic, loss_ucb);
- print_summary(signedrank_test_data2);
- """
- Wilcoxon signed-rank test
- W = 4.0, n = 44, median = -0.08, P = 7.96 × 10⁻¹³
- """
- # Sign-rank test Symbolic vs aUCB ------------------------------------------------------------
- signedrank_test_data2 = SignedRankTest(loss_symbolic, loss_aucb);
- print_summary(signedrank_test_data2);
- """
- Wilcoxon signed-rank test
- W = 22.0, n = 44, median = -0.05, P = 6.09 × 10⁻¹¹
- """
- # Combine the two datasets ------------------------------------------------------------
- loss_symbolic = vcat(
- # Data 1
- [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "Symbolic-model", "loo", "outcome",
- "gershman2018-data1", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:45
- ],
- # Data 2
- [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "Symbolic-model", "loo", "outcome",
- "gershman2018-data2", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:44
- ]
- );
- loss_ucb = vcat(
- [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "ucb-model", "loo", "outcome",
- "gershman2018-data1", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:45
- ],
- [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "ucb-model", "loo", "outcome",
- "gershman2018-data2", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:44
- ]
- );
- loss_aucb = vcat(
- [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "aucb-model", "loo", "outcome",
- "gershman2018-data1", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:45
- ],
- [
- mean(CSV.read(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "fit", "aucb-model", "loo", "outcome",
- "gershman2018-data2", "losses-$(string(subject)).csv"
- ), DataFrame
- ).loss)
- for subject in 1:44
- ]
- );
- begin
- fig = Figure(size = (600, 400));
- ax = GLMakie.Axis(fig[1, 1], ylabel = "Loss", xlabel = "Subject", limits = (0, 89, 0, 1));
- lines!(ax, loss_symbolic, label = "Symbolic");
- lines!(ax, loss_aucb, label = "UCB");
- axislegend(ax; position = :lt);
- fig
- end
- begin
- fig = Figure(size = (600, 400));
- ax = GLMakie.Axis(fig[1, 1], ylabel = "Loss", xlabel = "Subject");
- density!(ax, loss_symbolic .- loss_aucb)
- fig
- end
- # Sign-rank test Symbolic vs UCB ------------------------------------------------------------
- signedrank_test_data = SignedRankTest(loss_symbolic, loss_ucb);
- print_summary(signedrank_test_data);
- """
- Wilcoxon signed-rank test
- W = 15.0, n = 89, median = -0.07, P = 4.31 × 10⁻¹⁶
- """
- # Sign-rank test Symbolic vs aUCB ------------------------------------------------------------
- signedrank_test_data = SignedRankTest(loss_symbolic, loss_aucb);
- print_summary(signedrank_test_data);
- """
- Wilcoxon signed-rank test
- W = 286.0, n = 89, median = -0.03, P = 2.21 × 10⁻¹²
- """
- CairoMakie.activate!()
- begin
- fig = Figure(size = (450, 500), fontsize = 30);
- ax = GLMakie.Axis(
- fig[1, 1], ylabel = "Loss on Test Set", xlabel = "",
- limits = (-0.4, 1.4, 0.1, 0.9),
- xgridvisible = false, ygridvisible = false,
- xticks = (0:1, ["Symbolic", "intercept-UCB"]),
- yticks = 0.2:0.1:0.7,
- topspinevisible = false, rightspinevisible = false,
- xtrimspine = true, ytrimspine = true
- );
- x1 = 0 .+ (rand(89)*0.1.-0.05);
- x2 = 1 .+ (rand(89)*0.1.-0.05);
- [lines!(ax, [x1[i], x2[i]], [loss_symbolic[i], loss_aucb[i]], color=RGBA(0.8, 0.8, 0.8, 0.5)) for i in 1:89]
- scatter!(ax, x1, loss_symbolic, color=RGB(config["colors"]["ann"]...))
- scatter!(ax, x2, loss_aucb, color=RGB(config["colors"]["ucb"]...))
- lines!(ax, [0.2, 0.8], [mean(loss_symbolic), mean(loss_ucb)], color=RGBA(0, 0, 0, 1), linewidth = 2)
- scatter!(ax, [0.2], [mean(loss_symbolic)], color=RGB(config["colors"]["ann"]...), markersize = 20)
- scatter!(ax, [0.8], [mean(loss_ucb)], color=RGB(config["colors"]["ucb"]...), markersize = 20)
- display(fig)
- end
- CairoMakie.save(
- joinpath(
- config["mac"]["symbolic"], "cross-task-generalization",
- "figures", "compare-loss-aucb.png"
- ), fig, px_per_unit = 2
- )
- # Save figure
- [CairoMakie.save(
- joinpath(
- config["mac"]["figures"], "supplementary",
- "supplementary8.$ext"
- ), fig, px_per_unit = 2
- ) for ext in ["pdf", "png"]];
compare-loss.jl at commit 3d343be, under MIT · at the source
Overview
- Department of Experimental Psychology, University of Oxford,Oxford, UK
- Department of Psychology, New York University,New York, NY USA
- Nuffield Department of Clinical Neurosciences, University of Oxford,Oxford, UK
Abstract
Humans and other animals are driven to acquire information about opportunities in their environments, yet how they evaluate what is worth learning remains unclear. Here we combine artificial neural networks with symbolic regression to extract an expressive yet interpretable model that specifies how human participants evaluate decision-relevant information during choice. The recovered function depends primarily on the relative evidence accumulated across options rather than absolute uncertainty about each, revealing that participants seek information symmetry across alternatives rather than minimizing uncertainty option by option. This account outperforms standard models of uncertainty-based exploration and generalizes to an independent dataset. Using ultrahigh-field (7T) functional magnetic resonance imaging optimized for midbrain and brainstem, we simultaneously measured activity across five neuromodulatory nuclei and two cortical regions. Ventral tegmental area activity showed opposed coding of information and selection values, a pattern suited to arbitrating between sampling and choosing, and anterior cingulate cortex and anterior insula tracked value-of-information computations.
Reproduced under the paper's license (CC BY), from the paper cited above.
Repositories
Its files are read in the Code ↔ Paper reader above, with 17 matches between paragraphs and lines of code.
simonedambrogio/HybridModellingProject
3d343beb6dc7f1b7c6c7521ebec61f7dcef35b12, 25 June 2025Availability: 1 check, the latest on 27 September 2026: the link answers
- 27 September 2026: the link answers
141 files
- BehavioralModelling/
src/ , Julia, 63 linesattended-vs-unattended/ fit-model.jl - BehavioralModelling/
src/ , Julia, 69 linesattended-vs-unattended/ make-figure.jl - BehavioralModelling/
src/ , Julia, 9 linesfigure-config.jl - BehavioralModelling/
src/ , Julia, 69 linessample-vs-select/ fit-model.jl - BehavioralModelling/
src/ , Julia, 77 linessample-vs-select/ make-figure.jl - BehavioralModelling/
src/ , Julia, 115 linesspeed-accuracy/ subject-level/ make-figure.jl - BehavioralModelling/
src/ , Julia, 69 lines, 1 matchspeed-accuracy/ trial-level/ green-dots.jl - BehavioralModelling/
src/ , Julia, 176 linesspeed-accuracy/ trial-level/ make-figure.jl - BehavioralModelling/
src/ , Julia, 85 linesstay-vs-switch/ fit-model.jl - BehavioralModelling/
src/ , Julia, 79 linesstay-vs-switch/ make-figures.jl - BrainModelling/
first-level/ , Python, 110 linesscript/ feat/ all-visits.py - BrainModelling/
first-level/ , Python, 184 lines, 2 matchesscript/ feat/ create_design.py - BrainModelling/
first-level/ , Python, 62 linesscript/ feat/ fit.py - BrainModelling/
first-level/ , Python, 126 linesscript/ feat/ interaction-stay-switch. py - BrainModelling/
first-level/ , Python, 112 linesscript/ feat/ trial-level-intercept_fi xedtime.py - BrainModelling/
first-level/ , Python, 357 linesscript/ feat/ utils.py - BrainModelling/
first-level/ , Python, 23 linesscript/ post-feat/ mse/ combine.py - BrainModelling/
first-level/ , Python, 54 linesscript/ post-feat/ mse/ combine_and_save.py - BrainModelling/
first-level/ , Python, 71 linesscript/ post-feat/ mse/ compute.py - BrainModelling/
first-level/ , Julia, 194 linesscript/ post-feat/ mse/ roi-based-mse.jl - BrainModelling/
first-level/ , Python, 85 linesscript/ post-feat/ mse/ run_compute.py - BrainModelling/
first-level/ , Python, 70 linesscript/ post-feat/ mse/ run_compute_rois.py - BrainModelling/
first-level/ , Python, 80 linesscript/ post-feat/ mse/ run_standard2func.py - BrainModelling/
first-level/ , Python, 82 linesscript/ post-feat/ mse/ standard2func.py - BrainModelling/
first-level/ , Python, 93 linesscript/ post-feat/ mse/ visualize.py - BrainModelling/
first-level/ , Julia, 79 linesscript/ post-feat/ rsa/ fit-data/ load-data.jl - BrainModelling/
first-level/ , Julia, 142 linesscript/ post-feat/ rsa/ fit-data/ rsa.jl - BrainModelling/
first-level/ , Python, 388 linesscript/ post-feat/ rsa/ make-data/ create_masks.py - BrainModelling/
first-level/ , Python, 133 linesscript/ post-feat/ rsa/ make-data/ create_matrix.py - BrainModelling/
first-level/ , Python, 61 linesscript/ post-feat/ rsa/ make-data/ main.py - BrainModelling/
first-level/ , Shell, 85 linesscript/ post-feat/ rsa/ make-data/ submit.sh - BrainModelling/
first-level/ , Julia, 83 linesscript/ post-feat/ rsa/ make-figure.jl - BrainModelling/
first-level/ , Python, 70 linesscript/ post-feat/ symbolic-link.py - BrainModelling/
first-level/ , Shell, 41 linesscript/ submit.sh - BrainModelling/
second-level/ , Python, 3 linesscript/ feat/ check-inputs.py - BrainModelling/
second-level/ , Python, 63 linesscript/ feat/ create_design.py - BrainModelling/
second-level/ , Python, 32 linesscript/ feat/ fit.py - BrainModelling/
second-level/ , Python, 88 linesscript/ feat/ main.py - BrainModelling/
second-level/ , Julia, 245 lines, 1 matchscript/ post-feat/ roi/ fit-data/ fit-model.jl - BrainModelling/
second-level/ , Julia, 50 linesscript/ post-feat/ roi/ fit-data/ load-data.jl - BrainModelling/
second-level/ , Python, 62 linesscript/ post-feat/ roi/ make-data/ main.py - BrainModelling/
second-level/ , Python, 148 linesscript/ post-feat/ roi/ make-data/ make_dataframe.py - BrainModelling/
second-level/ , Julia, 261 linesscript/ post-feat/ roi/ make-figures.jl - BrainModelling/
second-level/ , Python, 187 linesscript/ utils.py - BrainModelling/
third-level/ , Python, 80 linesscript/ feat/ all-visits.py - BrainModelling/
third-level/ , Python, 65 linesscript/ feat/ create_design.py - BrainModelling/
third-level/ , Python, 31 linesscript/ feat/ fit.py - BrainModelling/
third-level/ , Python, 81 linesscript/ feat/ interacion-stay-switch.p y - BrainModelling/
third-level/ , Python, 68 linesscript/ feat/ main.py - BrainModelling/
third-level/ , Python, 162 linesscript/ feat/ utils.py - Data/
fit/ , Julia, 38 linesmake-data.jl - Data/
fsl/ , Shell, 13 linesdata/ masks/ 1mm/ convert2mm_1mm.sh - Data/
fsl/ , Shell, 30 linesdata/ masks/ 1mm/ erodeLC.sh - Data/
fsl/ , Shell, 23 linesdata/ masks/ 1mm/ newSN.sh - Data/
fsl/ , Shell, 5 linesdata/ masks/ 2mm/ binarize.sh - Data/
fsl/ , Shell, 13 linesdata/ masks/ 2mm/ convert1mm_2mm.sh - Data/
fsl/ , Shell, 8 linesdata/ masks/ make-ventral_Septal_Nucl ei.sh - Data/
fsl/ , Julia, 173 linesscr/ make_data.jl - Data/
fsl/ , Julia, 1,198 linesscr/ utils.jl - Data/
list/ , R, 628 linesscr/ get data_list.R - Data/
list/ , R, 31 linesscr/ get evs_fsl.R - Data/
list/ , R, 24 linesscr/ test.R - Data/
list/ , R, 94 linesscr/ utils.R - Data/
model/ , Julia, 75 linesmake_data.jl - Data/
voi/ , Julia, 100 linesmake_data.jl - HybridModelling/
src/ , Julia, 73 lines, 1 matchanalysis/ compare/ losses/ compare-losses.jl - HybridModelling/
src/ , Julia, 180 linesanalysis/ compare/ vois/ heatmap/ make-figure.jl - HybridModelling/
src/ , Julia, 23 linesanalysis/ compare/ vois/ helper.jl - HybridModelling/
src/ , Julia, 201 linesanalysis/ compare/ vois/ surfice/ make-figure.jl - HybridModelling/
src/ , Julia, 82 linesanalysis/ compare/ vois/ vector-field/ helper.jl - HybridModelling/
src/ , Julia, 300 linesanalysis/ compare/ vois/ vector-field/ make-figure.jl - HybridModelling/
src/ , Julia, 195 linesanalysis/ compare/ vois/ weights/ make-figure.jl - HybridModelling/
src/ , Julia, 156 lines, 1 matchanalysis/ fit/ cross-validation.jl - HybridModelling/
src/ , Julia, 203 linesanalysis/ fit/ estimation.jl - HybridModelling/
src/ , Shell, 13 linesanalysis/ fit/ hybrid-model/ cross-validation/ submit.sh - HybridModelling/
src/ , Julia, 132 linesanalysis/ fit/ hybrid-model/ estimation/ average-voi.jl - HybridModelling/
src/ , Julia, 56 linesanalysis/ fit/ hybrid-model/ estimation/ subject-voi.jl - HybridModelling/
src/ , Shell, 13 linesanalysis/ fit/ hybrid-model/ estimation/ submit.sh - HybridModelling/
src/ , Julia, 127 lines, 1 matchanalysis/ fit/ hybrid-model/ fit-hybrid.jl - HybridModelling/
src/ , Shell, 13 linesanalysis/ fit/ linear-model/ cross-validation/ submit.sh - HybridModelling/
src/ , Julia, 145 linesanalysis/ fit/ linear-model/ estimation/ average-voi.jl - HybridModelling/
src/ , Julia, 68 linesanalysis/ fit/ linear-model/ estimation/ subject-voi.jl - HybridModelling/
src/ , Shell, 13 linesanalysis/ fit/ linear-model/ estimation/ submit.sh - HybridModelling/
src/ , Julia, 60 linesanalysis/ fit/ linear-model/ fit-linear.jl - HybridModelling/
src/ , Shell, 19 linesanalysis/ fit/ symbolic-model/ cross-validation/ submit.sh - HybridModelling/
src/ , Julia, 145 linesanalysis/ fit/ symbolic-model/ estimation/ average-voi.jl - HybridModelling/
src/ , Julia, 63 linesanalysis/ fit/ symbolic-model/ estimation/ look-at-pars.jl - HybridModelling/
src/ , Julia, 68 linesanalysis/ fit/ symbolic-model/ estimation/ subject-voi.jl - HybridModelling/
src/ , Shell, 20 linesanalysis/ fit/ symbolic-model/ estimation/ submit-stanfit.sh - HybridModelling/
src/ , Shell, 18 linesanalysis/ fit/ symbolic-model/ estimation/ submit.sh - HybridModelling/
src/ , Julia, 82 linesanalysis/ fit/ symbolic-model/ fit-symbolic.jl - HybridModelling/
src/ , Julia, 165 lines, 1 matchanalysis/ fit/ symbolic-model/ symbolic2ann/ fit.jl - HybridModelling/
src/ , Julia, 451 lines, 1 matchanalysis/ fit/ symbolic-model/ symbolic2ann/ make-figure.jl - HybridModelling/
src/ , Shell, 12 linesanalysis/ fit/ ucb-model/ cross-validation/ submit.sh - HybridModelling/
src/ , Julia, 144 linesanalysis/ fit/ ucb-model/ estimation/ average-voi.jl - HybridModelling/
src/ , Julia, 68 linesanalysis/ fit/ ucb-model/ estimation/ subject-voi.jl - HybridModelling/
src/ , Shell, 13 linesanalysis/ fit/ ucb-model/ estimation/ submit.sh - HybridModelling/
src/ , Julia, 150 linesanalysis/ fit/ ucb-model/ fit-ucb.jl - HybridModelling/
src/ , Julia, 109 linesanalysis/ recover/ analyze/ correlation.jl - HybridModelling/
src/ , Julia, 326 linesanalysis/ recover/ debug.jl - HybridModelling/
src/ , Julia, 267 lines, 1 matchanalysis/ recover/ recover.jl - HybridModelling/
src/ , Julia, 135 lines, 1 matchanalysis/ symbolic-regression/ after-first-visit/ estimate.jl - HybridModelling/
src/ , Julia, 59 linesanalysis/ symbolic-regression/ after-first-visit/ make-input.jl - HybridModelling/
src/ , Julia, 18 linesanalysis/ symbolic-regression/ after-first-visit/ utils.jl - HybridModelling/
src/ , Julia, 300 lines, 2 matchesanalysis/ symbolic-regression/ cross-task-generalizatio n/ compare-loss.jl - HybridModelling/
src/ , Julia, 67 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ aucb-model/ fit-aucb.jl - HybridModelling/
src/ , Shell, 25 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ aucb-model/ loo/ submit.sh - HybridModelling/
src/ , Julia, 229 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ loo.jl - HybridModelling/
src/ , Julia, 102 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ models/ Symbolic.jl - HybridModelling/
src/ , Julia, 91 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ models/ UCB.jl - HybridModelling/
src/ , Julia, 61 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ models/ utils.jl - HybridModelling/
src/ , Julia, 27 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ recover.jl - HybridModelling/
src/ , Shell, 97 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ sub-loo.sh - HybridModelling/
src/ , Julia, 105 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ symbolic-model/ fit-symbolic.jl - HybridModelling/
src/ , Shell, 25 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ symbolic-model/ loo/ submit.sh - HybridModelling/
src/ , Julia, 97 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ ucb-model/ fit-ucb.jl - HybridModelling/
src/ , Shell, 25 linesanalysis/ symbolic-regression/ cross-task-generalizatio n/ fit/ ucb-model/ loo/ submit.sh - HybridModelling/
src/ , Julia, 231 lines, 1 matchanalysis/ symbolic-regression/ interpret/ make-figure.jl - HybridModelling/
src/ , Julia, 30 linesanalysis/ symbolic-regression/ robustness/ analyze.jl - HybridModelling/
src/ , Julia, 154 linesanalysis/ symbolic-regression/ robustness/ estimate.jl - HybridModelling/
src/ , Shell, 27 linesanalysis/ symbolic-regression/ robustness/ submit_estimate.sh - HybridModelling/
src/ , Julia, 360 linesmodel/ Hybrid.jl - HybridModelling/
src/ , Julia, 159 linesmodel/ Linear.jl - HybridModelling/
src/ , Julia, 217 linesmodel/ Symbolic.jl - HybridModelling/
src/ , Julia, 190 linesmodel/ UCB.jl - HybridModelling/
src/ , Julia, 178 linesmodel/ subjective-magnitude.jl - HybridModelling/
src/ , Julia, 1,461 linesutils.jl - OptimalPolicy/
src/ , Julia, 237 lines, 1 matchmdp.jl - OptimalPolicy/
src/ , Julia, 189 linessimulate/ debug.jl - OptimalPolicy/
src/ , Julia, 155 linessimulate/ evaluate-simulation.jl - OptimalPolicy/
src/ , Julia, 174 linessimulate/ helper.jl - OptimalPolicy/
src/ , Julia, 120 linessimulate/ simulate.jl - OptimalPolicy/
src/ , Julia, 105 lines, 2 matchessolve.jl - OptimalPolicy/
src/ , Shell, 39 linessubmit/ parse_args.sh - OptimalPolicy/
src/ , Shell, 21 linessubmit/ simulate.sh - OptimalPolicy/
src/ , Shell, 32 linessubmit/ solve.sh - OptimalPolicy/
src/ , Shell, 18 linessubmit/ solve_then_simulate.sh - OptimalPolicy/
src/ , Julia, 343 linesutils.jl - helper.jl, Julia, 74 lines
- LICENSE, License, 21 lines
- README.md, Text, 40 lines
Zenodo 19685085
Availability: 1 check, the latest on 27 September 2026: the link answers (HTTP 200)
- 27 September 2026: the link answers (HTTP 200)
Code availability
All analysis and modeling code is available via GitHub at https://
Reproduced under the paper's license (CC BY), from the paper cited above.
Tracing map
Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.
What the map holds:
- 2 repositories of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
- 139 scripts, each with its path and the digest of its content;
- 17 matches between paragraphs of the paper and lines of the code (method lexical-v1);
- neither the text of the paper nor the code itself.
Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.
Data
No dataset and no data link were found in the paper.
Data availability
All behavioral and neuroimaging data supporting the findings of this study are available via GitHub at https://
Reproduced under the paper's license (CC BY), from the paper cited above.
Versions
The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.
Version 2, 28 September 2026
- Publisher: n/a → Nature Portfolio
Version 1, 27 September 2026: the first record
Recorded: type, language, journal, volume, issue, pages, dates, 6 authors, 2 keywords, 13 MeSH terms, 2 funders, 68 references.
Cite
This paper
D’Ambrogio, S., Grohn, J., Khalighinejad, N., Mattar, M. G., Hunt, L., & Rushworth, M. F. S. (2026). Interpretable abstractions of artificial neural networks predict behavior and neural activity during human information gathering. Nature neuroscience, 29(8), 2036-2047. https://
BibTeX
@article{dambrogio2026in
author = {D’Ambrogio, Simone and Grohn, Jan and Khalighinejad, Nima and Mattar, Marcelo G. and Hunt, Laurence and Rushworth, Matthew F. S.},
title = {{Interpretable abstractions of artificial neural networks predict behavior and neural activity during human information gathering}},
journal = {Nature neuroscience},
year = {2026},
month = jun,
volume = {29},
number = {8},
pages = {2036--2047},
publisher = {Nature Portfolio},
issn = {1097-6256},
doi = {10.1038/
url = {https://
pmid = {42362882},
pmcid = {PMC13433322}
}
RIS
TY - JOUR
AU - D’Ambrogio, Simone
AU - Grohn, Jan
AU - Khalighinejad, Nima
AU - Mattar, Marcelo G.
AU - Hunt, Laurence
AU - Rushworth, Matthew F. S.
TI - Interpretable abstractions of artificial neural networks predict behavior and neural activity during human information gathering
T2 - Nature neuroscience
J2 - Nat Neurosci
PY - 2026
DA - 2026/
VL - 29
IS - 8
SP - 2036
EP - 2047
SN - 1097-6256
PB - Nature Portfolio
DO - 10.1038/
UR - https://
LA - en
ER -
CSL-JSON
{
"id": "10.1038/
"type": "article-journal",
"title": "Interpretable abstractions of artificial neural networks predict behavior and neural activity during human information gathering",
"container-title": "Nature neuroscience",
"author": [
{
"family": "D’Ambrogio",
"given": "Simone"
},
{
"family": "Grohn",
"given": "Jan"
},
{
"family": "Khalighinejad",
"given": "Nima"
},
{
"family": "Mattar",
"given": "Marcelo G."
},
{
"family": "Hunt",
"given": "Laurence"
},
{
"family": "Rushworth",
"given": "Matthew F. S."
}
],
"container-title-short":
"volume": "29",
"issue": "8",
"page": "2036-2047",
"DOI": "10.1038/
"PMID": "42362882",
"PMCID": "PMC13433322",
"ISSN": "1097-6256",
"publisher": "Nature Portfolio",
"URL": "https://
"language": "en",
"issued": {
"date-parts": [
[
2026,
6,
26
]
]
}
}
The tracing map gets a citation of its own once an author has validated it and it has a DOI.
Similar papers
The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.
- [1] doi:10.1073/pnas.2410955121 [code]
- A distributed subcortical circuit linked to instrumental information-seeking about threat.Journal: Proceedings of the National Academy of Sciences of the United States of AmericaIn common: data.table, seaborn, tidyverse, 4 other tools, 16 references, 2 authors
- [2] doi:10.1038/s41467-026-68349-9 [code]
- Activity in human dorsal raphe nucleus signals changes in behavioural policy.Journal: Nature communicationsIn common: tidyverse, 18 references, 2 authors
- [3] doi:10.1038/s41467-024-49128-w [code]
- General mechanisms of task engagement in the primate frontal cortex.Journal: Nature communicationsIn common: 7 references, 3 authors
- [4] doi:10.1038/s41562-026-02473-w [code]
- Brain activity, disruption and connectivity comparisons identify origins of human metacognition in other primates.Journal: Nature human behaviourIn common: FSL, 4 references, 3 authors
- [5] doi:10.1038/s41467-025-65766-0 [code]
- Interaction and functional specialization across a distributed neural circuit for flexible task control in macaques.Journal: Nature communicationsIn common: 6 references, 2 authors
- [6] doi:10.1038/s41467-025-62507-1 [code]
- Higher-order and distributed synergistic functional interactions encode information gain in goal-directed learning.Journal: Nature communicationsIn common: NiBabel, pandas, SciPy, 2 other tools, 8 references
- [7] doi:10.1038/s41562-024-01971-z [code]
- Common neural choice signals can emerge artefactually amid multiple distinct value signals.Journal: Nature human behaviourIn common: Makie, DataFrames.jl, Distributions.jl, 4 other tools, 3 references
- [8] doi:10.1038/s41467-025-67853-8 [code]
- Active disambiguation guides inferring controllability and cause in social interactions.Journal: Nature communicationsIn common: tidyverse, 7 references, author Matthew F. S. Rushworth
- [9] doi:10.1038/s41467-024-55459-5 [code]
- Network structure influences the strength of learned neural representations.Journal: Nature communicationsIn common: Makie, DataFrames.jl, Distributions.jl, 8 other tools, 1 reference
- [10] doi:10.1073/pnas.2506243122
- Fundamental features of social environments determine rate of social affiliation.Journal: Proceedings of the National Academy of Sciences of the United States of AmericaIn common: 7 references, author Matthew F. S. Rushworth
Contribute
The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.
Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.
Claim this paper
Correct its record
Say what each link of this record is, remove the ones that are not the paper's, add the ones that are missing. The correction becomes a new version of the record, in its Versions section.
Validate its tracing map
You validate the map as this page shows it: 2 repositories of the authors' code, each at its verified commit and with its license, 139 scripts, and 17 matches between paragraphs and code (see the Code and Map sections). It then receives a DOI on Zenodo, with you (your ORCID iD) and OSCR as its creators; the code itself is not deposited.
The map's fingerprint: sha256:6a9315e7974347ee…
Add the badge to the code's README
If you are an author or a maintainer of this paper's code, you can add this badge. It says what OSCR knows of the paper (its code, its validated tracing map) and links the code to this page. Paste one of these snippets under the README's title. Keep the two marks around it, so that the badge can be replaced in place and is never added twice. OSCR changes nothing in your repository.
Markdown
<!-- oscr-badge:start -->
[](https://openscicode.org/badge/doi/10.1038/s41593-026-02342-9)
<!-- oscr-badge:end -->reStructuredText
.. oscr-badge:start
.. image:: https://openscicode.org/badge/doi/10.1038/s41593-026-02342-9.svg
:target: https://openscicode.org/badge/doi/10.1038/s41593-026-02342-9
:alt: OSCR
.. oscr-badge:endHTML
<!-- oscr-badge:start -->
<a href="https://openscicode.org/badge/doi/10.1038/s41593-026-02342-9"><img src="https://openscicode.org/badge/doi/10.1038/s41593-026-02342-9.svg" alt="OSCR"></a>
<!-- oscr-badge:end -->If you choose to, you can have the badge kept up to date for you, with oscr badge --write from the command line or with the GitHub Action in your repository's workflows. Every other form and option (a card, AsciiDoc, LaTeX, shields.io) is on the badge page.
Request its removal
To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).
Discussion, reproductions, activity
Discussion: code errors and mismatches between the code and the paper about this paper and its code, from readers and its authors, beside the paper and its tracing map.
Report a mismatch between the code and the paper · Report a code error (signed in)
Reproductions: reports from readers who ran the authors' code and did not get the paper's result, what they ran, with which environment, commit and data.
Report a reproduction failure (signed in)
Versions of its code: the releases of its code tied to a version of this paper (the preprint, the accepted manuscript, the version of record…), each with the tracing map versioned with it.
Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.
