Upload folder using huggingface_hub

README.md

@@ -0,0 +1,70 @@
+---
+pretty_name: BeliefSim
+language:
+- en
+task_categories:
+- text-classification
+- tabular-classification
+tags:
+- misinformation
+- belief-modeling
+- survey-data
+- demographic-analysis
+license: other
+---
+
+# BeliefSim Dataset Release
+
+This folder contains an anonymized, Hugging Face-ready data package for **BeliefSim: Towards Belief-Driven Simulation of Demographic Misinformation Susceptibility**.
+
+## Content Note
+
+This dataset studies real misinformation and rumor examples. Some claims may contain offensive, stigmatizing, or otherwise harmful wording. Examples are included for research transparency and should be handled with care.
+
+## Files
+
+- `data/claims.csv`: claim text from PANDORA and MIST-1, with source labels and gold labels where available.
+- `data/judgments.csv`: anonymized participant-level claim judgments.
+- `data/evaluation_instances.csv`: target claim judgments paired with two same-participant observed belief judgments.
+- `data/wvs_group_priors.csv`: WVS-derived group belief distributions by demographic axis.
+- `data/dataset_summary.json`: generated counts and privacy notes.
+- `scripts/prepare_release.py`: reproducible script used to regenerate the release files from local raw sources.
+
+## Current Generated Counts
+
+- Claims: 156 total, including 56 PANDORA and 100 MIST-1 claims.
+- Judgments: 40,964 total, including 64 PANDORA and 40,900 MIST-1 judgments.
+- Evaluation instances: 40,135 total, including 53 PANDORA and 40,082 MIST-1 instances.
+- WVS group prior rows: 976.
+
+These counts are generated from the local files available in this workspace at preparation time.
+
+## Anonymization
+
+The release script removes direct identifiers and sensitive survey metadata before writing the public CSV files. In particular, it drops response IDs, Prolific IDs, IP addresses, latitude/longitude, raw timestamps, and free-text notes/comments. Participant identifiers are replaced with deterministic salted SHA-256 hashes. Demographic fields are kept only as broad categories such as gender, age group, education bucket, and living-area group where available.
+
+Before public upload, verify that the upstream PANDORA, MIST-1, and WVS data-use terms permit redistribution of the fields included here. If redistribution is not permitted for any source, upload only the scripts/metadata and provide instructions for authorized users to reconstruct the dataset locally.
+
+## Loading
+
+```python
+from datasets import load_dataset
+
+claims = load_dataset("MichiganNLP/beliefsim", data_files="data/claims.csv")
+judgments = load_dataset("MichiganNLP/beliefsim", data_files="data/judgments.csv")
+instances = load_dataset("MichiganNLP/beliefsim", data_files="data/evaluation_instances.csv")
+wvs_priors = load_dataset("MichiganNLP/beliefsim", data_files="data/wvs_group_priors.csv")
+```
+
+## Citation
+
+```bibtex
+@article{borah2026belief,
+  title={Belief-Sim: Towards Belief-Driven Simulation of Demographic Misinformation Susceptibility},
+  author={Borah, Angana and Khan, Zohaib and Mihalcea, Rada and P{\'e}rez-Rosas, Ver{\'o}nica},
+  journal={arXiv preprint arXiv:2603.03585},
+  year={2026}
+}
+```
+
+Questions or collaboration inquiries about misinformation research can be directed to anganab@umich.edu.

UPLOAD_STEPS.md

@@ -0,0 +1,40 @@
+# Uploading BeliefSim to Hugging Face
+
+Run these commands from `/home/anganab/misinfo` after reviewing the generated files and confirming upstream redistribution permissions.
+
+```bash
+cd /home/anganab/misinfo
+python3 beliefsim_hf_dataset/scripts/prepare_release.py
+```
+
+Install the Hugging Face CLI if needed:
+
+```bash
+python3 -m pip install --user -U huggingface_hub
+```
+
+Log in with a Hugging Face token that has write access:
+
+```bash
+huggingface-cli login
+```
+
+Create the dataset repository. Use the org name if you want it under MichiganNLP:
+
+```bash
+huggingface-cli repo create MichiganNLP/beliefsim --type dataset --private
+```
+
+Upload the prepared folder:
+
+```bash
+huggingface-cli upload MichiganNLP/beliefsim beliefsim_hf_dataset . --repo-type dataset
+```
+
+Open the dataset page, inspect the preview, then switch visibility to public only after verifying that PANDORA, MIST-1, and WVS redistribution terms permit the included fields. If permissions are uncertain, keep the repo private or upload only `README.md` and `scripts/prepare_release.py`.
+
+Suggested public URL after upload:
+
+```text
+https://huggingface.co/datasets/MichiganNLP/beliefsim
+```

data/claims.csv

@@ -0,0 +1,157 @@
+source,claim_id,claim_text,gold_label,content_warning
+PANDORA,Q256,Police confirm that #sydneysiege is finally over. Two people reportedly injured after suffering gunshot wounds http://t.co/nsprnSYTDT,,True
+PANDORA,Q259,"Police have clarified that there were two shootings in Ottawa today, not three: at the War Memorial and Parliament Hill.",,True
+PANDORA,Q262,Police have surrounded this building where the suspected #CharlieHebdo attackers are holed-up http://t.co/B0TRvE2iYs http://t.co/becxNG0Kxi,,True
+PANDORA,Q266,"Police say shots fired at 3 #Ottawa sites - National War Memorial, Parliament Hill, and now Rideau shopping centre http://t.co/rOnqpUbuqc",,True
+PANDORA,Q269,Police taking people out of #Paris grocery where hostages were being held. http://t.co/7Bxdhozq4H,,True
+PANDORA,Q272,"Prince not playing Massey Hall tonight, promoter says http://t.co/clVHvM07kx http://t.co/IADirsFOOC",,True
+PANDORA,Q275,"Ray Hadley says he spoke with hostage, and could hear the gunman in the background barking orders and demanding to go live on air",,True
+PANDORA,Q278,Reports on @channeltennews #sydneysiege gunman wants #ISIL flag delivered to cafe and he also wants to speak to PM or people will be killed.,,True
+PANDORA,Q281,SYDNEY SIEGE: Gunman forces hostages to hold up ISIS flag in window http://t.co/9K3XXEiSFa http://t.co/cATxCjYXpy,,True
+PANDORA,Q284,Secret Service tells Infowars: Hillary seriously ill; needs special accommodations Tune in 4pm-6pm CST https://t.co/OQtch0Lewb #trump #tcot,,True
+PANDORA,Q287,Secret Service tells Infowars: Hillary seriously ill; needs special accommodations Tune in 4pm-6pm CST https://t.co/OQtch0Lewb #trump #tcot,,True
+PANDORA,Q290,Senior U.S. official: Canadian government has informed U.S. that one shooter is dead in Ottawa. Live blog: http://t.co/q98AMohu7T,,True
+PANDORA,Q293,Several hostages freed at Jewish supermarket in Paris. Photo Thomas Samson #AFP http://t.co/C9ltgMT0Wl,,True
+PANDORA,Q296,Soldier killed in Canada shooting was a young reservist with a six-year-old son http://t.co/83PReIl9XV http://t.co/LVprXOCCF7,,True
+PANDORA,Q299,Sydney airspace wasn't closed. A second terror suspect wasn't arrested. Myths around #sydneysiege debunked. http://t.co/yTLKLUKnFd,,True
+PANDORA,Q302,The Sydney cafe siege may be part of a larger plot. @Y7News #MartinPlaceSiege http://t.co/f2aTNEpjmh http://t.co/DE4IBxv73K,,True
+PANDORA,Q305,The gunman in Ottawa has been shot and killed. I'm at a loss for words this morning. That isn't my Canada.,,True
+PANDORA,Q308,"The latest @CBCNews on #Ottawa shootings. Gunman killed inside Parliament, soldier shot at War Memorial. http://t.co/kngapKTSCe #cdnpoli",,True
+PANDORA,Q311,These are not timid colours; soldiers back guarding Tomb of Unknown Soldier after today's shooting #StandforCanada http://t.co/7KoW2xATKG,,True
+PANDORA,Q314,This afternoon we've lowered our flags to half mast in honour of the Canadian Reservist who lost his life in Ottawa. http://t.co/3oTF5sd2Lf,,True
+PANDORA,Q317,This is crazy! #CapeTown #capestorm #weatherforecast https://t.co/3bcKOKrCJB,,True
+PANDORA,Q320,"UPDATE: 12 now confirmed dead in #CharlieHebdo attack in Paris, @AP is reporting. Story: http://t.co/Jfqmha1GdV (AP) http://t.co/vDCIFHIa9U",,True
+PANDORA,Q323,"UPDATE: 13 people being held hostage in #Sydney shop, Opera House evacuated after suspicious package found: reports http://t.co/n4D3yGjso9",,True
+PANDORA,Q326,Uber Sydney trips from CBD will be free for riders. Higher rates are still in place to encourage drivers to get into the CBD.,,True
+PANDORA,Q329,Uber intros surge pricing in downtown Sydney during hostage siege http://t.co/XWGe9MpUzU http://t.co/cwhIiozFNr,,True
+PANDORA,Q332,"Uber says it has hiked prices in Sydney to ""encourage more drivers to come online & pick up passengers."" http://t.co/MCJ0rMSXPc #SydneySiege",,True
+PANDORA,Q335,Unconfirmed reports claim that Michael Essien has contracted Ebola. http://t.co/VASQrZdLhH,,True
+PANDORA,Q338,Unformed Russian Embassy staff in London have left for Russia Rumours Putin HAS DIED! http://t.co/zSIV8w6FJ2 via @ShaunyNews #PutinDead?,,True
+PANDORA,Q341,"WTF RT @Jeggit: Believe it or not, this is a shark on the freeway in Houston, Texas. #HurricaneHarvy https://t.co/Vpikg0zguV",,True
+PANDORA,Q344,Watch video showing gunfire inside Canada's parliament in Ottawa http://t.co/CJpXNAk8nS http://t.co/hxwr2NEr2K,,True
+PANDORA,Q347,“@LePoint: #CharlieHebdo : “The cartoonists Charb & cabu are dead.” http://t.co/7USD83cQd2 http://t.co/WGHbLiLFoX”,,True
+PANDORA,Q350,“@LePoint: #CharlieHebdo : “The cartoonists Charb & cabu are dead.” http://t.co/7USD83cQd2 http://t.co/WGHbLiLFoX”,,True
+PANDORA,Q353,R_A Desperate Scramble as Covid-19 Families vie for Access to Plasma Therapy,,True
+PANDORA,Q356,R_Adults with Covid-19 Twice as Likely to Have Eaten at Restaurants,,True
+PANDORA,Q359,R_Amazon Plans to Prosecute Sellers for Coronavirus Outbreak.jpg,,True
+PANDORA,Q362,"R_American Incomes Were Raising, Until the Pandemic Hit",,True
+PANDORA,Q365,"R_Amid Coronavirus Outbreak, Carnival Cruise Line Offers on-Ship Credits to Passengers Who Don't Reschedule.jpeg",,True
+PANDORA,Q368,R_Anti-Inflammatory Drug may Shorten Covid-19 Recovery Time,,True
+PANDORA,Q371,R_AstraZeneca Resumes Its COVID-19 Vaccine Trials in the U.K.,,True
+PANDORA,Q374,"R_Coronavirus can Hijack Brain Cells to Replicate Itself, Yale Researchers Discover",,True
+PANDORA,Q377,R_Coronavirus Poses Tough Challenge for Economic Policymakers.jpeg,,True
+PANDORA,Q380,R_Coronavirus Vaccine Showing Promise in Early Human Trials,,True
+PANDORA,Q383,R_Covid-19 Cases Among Florida Children Jumped 26 Percent in a Month.,,True
+PANDORA,Q386,"R_CDC Coronavirus Spread may Last into 2021, but Impact can be Blunted.jpeg",,True
+PANDORA,Q389,R_DOJ Seeks New Emergency Powers Amid Coronavirus Pandemic,,True
+PANDORA,Q392,R_Experimental Drug Holds Promise for Treating the Coronavirus,,True
+PANDORA,Q395,R_Fauci Says It's Doable to have Millions of Doses of Covid-19 Vaccine by January,,True
+PANDORA,Q398,R_Germany Boosts Investment in Covid-19 Vaccine Research,,True
+PANDORA,Q401,R_Israel Declares 14 Day Quarantine for all Arrivals.jpeg,,True
+PANDORA,Q404,R_It's Irresponsible and It's Dangerous-Experts Rip Trump's Idea of Injecting Disinfectant to Treat COVID-19,,True
+PANDORA,Q407,R_N.I.H. Director has Cautious Optimism for Covid-19 Vaccine by End of 2020,,True
+PANDORA,Q410,R_Ohio Man Dies from COVID-19 After Criticizing Governor's Coronavirus Lockdown as Madness.jpg,,True
+PANDORA,Q413,"R_Pennsylvania Coronavirus Restrictions Deemed Unconstitutional, Federal Judge Rules",,True
+PANDORA,Q416,R_Police in the US Spread a False Claim that Meth is Contaminated with Coronavirus.jpeg,,True
+PANDORA,Q419,R_Some Urban Hospitals Face Closure or Cutback as the Pandemic Adds to Fiscal Woes,,True
+PANDORA,News_1_share,No,,True
+MIST-1,MIST_1,A Small Group of People Control the World Economy by Manipulating the Price of Gold and Oil,misinformation,True
+MIST-1,MIST_2,A Study Found That the Mainstream Media Is Lying about the Cause of Lyme Disease,misinformation,True
+MIST-1,MIST_3,African Nations Are Expected to Hand Over Control of Their Resources to Western Investors Within a Decade,misinformation,True
+MIST-1,MIST_4,Black People Are More Likely to Be Criminals,misinformation,True
+MIST-1,MIST_5,Breaking News: Australia Is the Only Country That Lets Pedophiles Off the Hook,misinformation,True
+MIST-1,MIST_6,Breaking News: Scientists Found a Link Between Climate Change and Tooth Decay,misinformation,True
+MIST-1,MIST_7,Certain Vaccines Are Loaded with Dangerous Chemicals and Toxins,misinformation,True
+MIST-1,MIST_8,"Climate Scientists' Work Is 'Unreliable', a 'Deceptive Method of Communication'",misinformation,True
+MIST-1,MIST_9,"Doctors Are Paid to Promote All Kinds of Ridiculous, Harmful, and Dangerous Treatments",misinformation,True
+MIST-1,MIST_10,"Ebola Virus 'Caused by US Nuclear Weapons Testing', New Study Says",misinformation,True
+MIST-1,MIST_11,Fact: Left-Wing/Left-Wing Media Exaggerates the 'White Flight' of West Coast Whites Towards Minority Enclaves,misinformation,True
+MIST-1,MIST_12,Google Has Made It Illegal to Film Police Officers in the US,misinformation,True
+MIST-1,MIST_13,Government Officials Have Illegally Manipulated the Weather to Cause Devastating Storms,misinformation,True
+MIST-1,MIST_14,Government Officials Have Manipulated Stock Prices to Hide Scandals,misinformation,True
+MIST-1,MIST_15,"Left-Wing Extremism Causes 'More Damage' to World Than Terrorism, Says UN Report",misinformation,True
+MIST-1,MIST_16,Left-Wingers Are More Likely to Lie to Get a Good Grade,misinformation,True
+MIST-1,MIST_17,New Data Show Shark Fins Are the 'Most Important Source of Protein' for the World's Poor,misinformation,True
+MIST-1,MIST_18,New Study: Americans Are the Most Religious People on Earth,misinformation,True
+MIST-1,MIST_19,New Study: Clear Relationship Between Eye Color and Intelligence,misinformation,True
+MIST-1,MIST_20,New Study: Left-Wingers Are More Likely to Lie to Get a Higher Salary,misinformation,True
+MIST-1,MIST_21,New Study: Liberals Are More Likely to Be Overweight,misinformation,True
+MIST-1,MIST_22,"Right-Wing Extremists: 5% of Population, 20% of Voting Power, 30% of the Wealth, 70% of the Police and Army",misinformation,True
+MIST-1,MIST_23,Science Proves That Conservatives Are More Violent than Liberals,misinformation,True
+MIST-1,MIST_24,Scientists Are Working to Genetically Engineer Food to Produce a Toxin That Will Give It Aphrodisiac Qualities,misinformation,True
+MIST-1,MIST_25,Scientists at Harvard Are Now Talking about the Possibility of the Moon Landing Being Faked,misinformation,True
+MIST-1,MIST_26,Scientists Find 'Genetic Basis' for Right-Wing Extremist Views,misinformation,True
+MIST-1,MIST_27,"Sudden Infant Death Syndrome (SIDS) Is Now a ""Scientific"" Epidemic, and the Media and Medical Community Has Been Covering It up for Decades",misinformation,True
+MIST-1,MIST_28,The Corporate Media Is Controlled by the Military-industrial Complex: The Major Oil Companies Own the Media and Control Their Agenda,misinformation,True
+MIST-1,MIST_29,The Drug Companies Have Been Controlling the Political Arena for Over a Hundred Years,misinformation,True
+MIST-1,MIST_30,"The Government Has Secret Information on Every Citizen, Including Their Financial and Medical Information",misinformation,True
+MIST-1,MIST_31,The Government Is Actively Destroying Evidence Related to the JFK Assassination,misinformation,True
+MIST-1,MIST_32,The Government Is Conducting a Massive Cover-Up of Their Involvement in 9/11,misinformation,True
+MIST-1,MIST_33,The Government Is Knowingly Spreading Disease Through the Airwaves and Food Supply,misinformation,True
+MIST-1,MIST_34,The Government Is Manipulating the Public's Perception of Genetic Engineering in Order to Make People More Accepting of Such Techniques,misinformation,True
+MIST-1,MIST_35,"The Government Is Secretly Stockpiling Food and Other Goods in Case of a ""Hard-Currency Collapse""",misinformation,True
+MIST-1,MIST_36,"The Mainstream Media Is Misinforming about Vaccinations: Vaccines Cause Autism, Vaccines Are Unsafe and Dangerous",misinformation,True
+MIST-1,MIST_37,"The More You Learn About the Moon Landing, the More You Realize That It Was Faked",misinformation,True
+MIST-1,MIST_38,"The So-Called ""Overdose"" Epidemic Is a Myth, and Those Who Believe It Are ""Nutritional Extremists""",misinformation,True
+MIST-1,MIST_39,UFO Sightings Are Actually the Result of Secret Military Experiments,misinformation,True
+MIST-1,MIST_40,UK Will Have to Spend up to £23bn on Military in Response to Right-Wing Extremism,misinformation,True
+MIST-1,MIST_41,UN Report: Worldwide Rise of Infant Mortality Due to Modern Medicine,misinformation,True
+MIST-1,MIST_42,UN: More than 50% of the World's Population Will Face Famine by 2070,misinformation,True
+MIST-1,MIST_43,UN: New Report Shows Shark Fin Soup as 'the Most Important Source of Protein' for World's Poor,misinformation,True
+MIST-1,MIST_44,US Is the Most Dangerous Place on Earth for Women,misinformation,True
+MIST-1,MIST_45,64% of Americans Want Stricter Laws on Gun Sales,true_information,True
+MIST-1,MIST_46,83% Say Measles Vaccine Is Safe for Healthy Children - No Partisan Differences in Views of Vaccine Safety,true_information,True
+MIST-1,MIST_47,About a Quarter of Large US Newspapers Laid off Staff in 2018,true_information,True
+MIST-1,MIST_48,Americans Feel Generally Positive about Their Own Finances,true_information,True
+MIST-1,MIST_49,"Americans Have Positive Views about Religion’s Role in Society, But Want It out of Politics",true_information,True
+MIST-1,MIST_50,"Attitudes Toward EU Are Largely Positive, Both Within Europe and Outside It",true_information,True
+MIST-1,MIST_51,Blue Light Isn’t the Main Source of Eye Fatigue and Sleep Loss - It’s Your Computer,true_information,True
+MIST-1,MIST_52,Data Show Upswing in Child Exploitation Cases,true_information,True
+MIST-1,MIST_53,Democrats More Supportive than Republicans of Federal Spending for Scientific Research,true_information,True
+MIST-1,MIST_54,"East Germany Has Narrowed Economic Gap with West Germany Since Fall of Communism, But Still Lags",true_information,True
+MIST-1,MIST_55,Experts Optimistic About the Next 50 Years of Digital Life,true_information,True
+MIST-1,MIST_56,Four in 10 Teachers Say Their School Is Not Well Protected,true_information,True
+MIST-1,MIST_57,Germany Welcomes Tesla's Berlin Move as Boost to E-Car Shift,true_information,True
+MIST-1,MIST_58,Global Warming Age Gap: Younger Americans Most Worried,true_information,True
+MIST-1,MIST_59,Hawaii May Subpoena Airbnb for Tax Records of Rental Hosts,true_information,True
+MIST-1,MIST_60,Hyatt Will Remove Small Bottles from Hotel Bathrooms by 2021,true_information,True
+MIST-1,MIST_61,"In Emerging Economies, Smartphone and Social Media Users Have Broader Social Networks",true_information,True
+MIST-1,MIST_62,"In US, 54% Say Teachers Unprepared to Handle Discipline",true_information,True
+MIST-1,MIST_63,"In US, 71% Support Transgender People Serving in Military",true_information,True
+MIST-1,MIST_64,"In Western European Countries with Church Taxes, Support for the Tradition Remains Strong",true_information,True
+MIST-1,MIST_65,International Relations Experts and US Public Agree: America Is Less Respected Globally,true_information,True
+MIST-1,MIST_66,Is US Fertility at An All-time Low? Two of Three Measures Point to Yes,true_information,True
+MIST-1,MIST_67,Japan's Ageing 'Hidden Christians' Fear They May Be Their Religion's Last Generation,true_information,True
+MIST-1,MIST_68,"Latinos Are More Likely to Believe in the American Dream, But Most Say It Is Hard to Achieve",true_information,True
+MIST-1,MIST_69,Leaked Russian Interference Report Raises UK Vote Questions,true_information,True
+MIST-1,MIST_70,"Majority in US Still Want Abortion Legal, with Limits",true_information,True
+MIST-1,MIST_71,Minnesota School District May Monitor Students’ Online Posts,true_information,True
+MIST-1,MIST_72,More People Around the World See US Power and Influence as a ‘Major Threat’ to Their Country,true_information,True
+MIST-1,MIST_73,Morocco’s King Appoints Committee Chief to Fight Poverty and Inequality,true_information,True
+MIST-1,MIST_74,Most Americans Say It’s OK for Professional Athletes to Speak out Publicly about Politics,true_information,True
+MIST-1,MIST_75,Most Americans Say Science Has Brought Benefits to Society and Expect More to Come,true_information,True
+MIST-1,MIST_76,Most Distant World Ever Explored Gets New Name: Arrokoth,true_information,True
+MIST-1,MIST_77,National Politics on Twitter: Small Share of US Adults Produce Majority of Tweets,true_information,True
+MIST-1,MIST_78,Nearly Half in US Fear Being the Victim of a Mass Shooting,true_information,True
+MIST-1,MIST_79,Nearly Six-in-Ten Americans Say Abortion Should Be Legal in All or Most Cases,true_information,True
+MIST-1,MIST_80,Norway's Oil Decline Accelerates,true_information,True
+MIST-1,MIST_81,"Obama Tops Public’s List of Best President in Their Lifetime, Followed by Clinton, Reagan",true_information,True
+MIST-1,MIST_82,One-in-Three Worldwide Lack Confidence in NGOs,true_information,True
+MIST-1,MIST_83,One-in-Five Americans Now Listen to Audiobooks,true_information,True
+MIST-1,MIST_84,"One-in-Five US Newsroom Employees Live in New York, Los Angeles or DC",true_information,True
+MIST-1,MIST_85,Online Grocery Shopping Still Rare in US,true_information,True
+MIST-1,MIST_86,Political Typology Reveals Deep Fissures on the Right and Left,true_information,True
+MIST-1,MIST_87,"Reflecting a Demographic Shift, 109 US Counties Have Become Majority Nonwhite Since 2000",true_information,True
+MIST-1,MIST_88,"Republicans Divided in Views of Trump’s Conduct, Democrats Are Broadly Critical",true_information,True
+MIST-1,MIST_89,"Republicans, Democrats See Opposing Party as More Ideological Than Their Own",true_information,True
+MIST-1,MIST_90,"Since Trump’s Election, Increased Attention to Politics – Especially Among Women",true_information,True
+MIST-1,MIST_91,Students at Smaller Colleges More Likely to Say Faculty Care,true_information,True
+MIST-1,MIST_92,Taiwan Seeks to Join Fight Against Global Warming,true_information,True
+MIST-1,MIST_93,Two in Three Americans Now Support Legalizing Marijuana,true_information,True
+MIST-1,MIST_94,United Nations Gets Mostly Positive Marks from People Around the World,true_information,True
+MIST-1,MIST_95,"Universal Basic Income Favored in Canada, UK But Not in US",true_information,True
+MIST-1,MIST_96,"US Hispanic Population Reached New High in 2018, But Growth Has Slowed",true_information,True
+MIST-1,MIST_97,US Panel Backs Wider Use of Fish Oil to Prevent Heart Attack,true_information,True
+MIST-1,MIST_98,US Safety Board Wants Boeing to Redesign 737 NG Part after Fatal Southwest Accident,true_information,True
+MIST-1,MIST_99,US Support for Legal Marijuana Steady in Past Year,true_information,True
+MIST-1,MIST_100,"Volkswagen to Invest 60 Billion Euros in E-mobility, Digital Tech by 2024",true_information,True

data/dataset_summary.json

@@ -0,0 +1,31 @@
+{
+  "claims": 156,
+  "judgments": 40964,
+  "evaluation_instances": 40135,
+  "wvs_group_prior_rows": 976,
+  "sources": {
+    "PANDORA": {
+      "claims": 56,
+      "judgments": 64,
+      "evaluation_instances": 53
+    },
+    "MIST-1": {
+      "claims": 100,
+      "judgments": 40900,
+      "evaluation_instances": 40082
+    }
+  },
+  "privacy": {
+    "dropped": [
+      "ResponseId",
+      "PROLIFIC_ID",
+      "IPAddress",
+      "LocationLatitude",
+      "LocationLongitude",
+      "open-text notes/comments",
+      "raw timestamps"
+    ],
+    "participant_ids": "deterministic salted SHA-256 hashes, truncated to 16 hex characters",
+    "demographics": "broad gender, age bucket, education bucket, and living-area fields only"
+  }
+}

scripts/prepare_release.py

@@ -0,0 +1,437 @@
+#!/usr/bin/env python3
+"""Prepare an anonymized BeliefSim dataset package for Hugging Face.
+
+The script keeps claims, participant judgments, evaluation triplets, and WVS
+group priors. It drops direct identifiers, IP/location metadata, and open-text
+free-response notes.
+"""
+
+import hashlib
+import json
+import re
+import zipfile
+from pathlib import Path
+from xml.etree import ElementTree as ET
+
+import pandas as pd
+
+
+ROOT = Path(__file__).resolve().parents[2]
+OUT = Path(__file__).resolve().parents[1] / "data"
+SALT = "beliefsim-release-v1"
+
+
+PANDORA_PATH = ROOT / "misinfo_combined_latest_with_acc_new.csv"
+MIST_PATH = ROOT / "wvs_misinfo" / "MIST - Sample 1 - Raw Dataset.csv"
+MIST_ITEM_DB = ROOT / "wvs_misinfo" / "MIST - Phase 4 - Item Database (January 2020).xlsx"
+
+WVS_FILES = {
+    "gender": ROOT / "wvs_misinfo" / "gender_alldimensions_wvs.csv",
+    "living_area": ROOT / "wvs_misinfo" / "urbrur_alldimensions_wvs.csv",
+    "age": ROOT / "wvs_misinfo" / "age_alldimensions.csv",
+    "education": ROOT / "wvs_misinfo" / "education_alldimensions.csv",
+}
+
+
+def anonymize(prefix: str, *parts: object) -> str:
+    raw = "|".join("" if p is None else str(p) for p in parts)
+    digest = hashlib.sha256(f"{SALT}|{prefix}|{raw}".encode("utf-8")).hexdigest()[:16]
+    return f"{prefix}_{digest}"
+
+
+def clean_text(value: object) -> str:
+    if pd.isna(value):
+        return ""
+    return re.sub(r"\s+", " ", str(value)).strip()
+
+
+def norm(value: object) -> str:
+    return re.sub(r"[^a-z0-9]+", " ", clean_text(value).lower()).strip()
+
+
+def first_nonempty(row: pd.Series, candidates: list) -> str:
+    for col in candidates:
+        if col in row.index:
+            val = clean_text(row[col])
+            if val and val.lower() != "nan":
+                return val
+    return ""
+
+
+def extract_claim(prompt_text: str) -> str:
+    text = re.sub(r"^\s*News:\s*", "", clean_text(prompt_text), flags=re.I)
+    for marker in ["Supporting Stance:", "Refuting Stance:", "Your responses:", "Your response:"]:
+        if marker in text:
+            text = text.split(marker, 1)[0]
+    return clean_text(text)
+
+
+def standardize_pandora_judgment(value: object) -> str:
+    v = norm(value)
+    if not v:
+        return ""
+    if "importid" in v or v in {"true information", "misinformation", "have you heard of the information before", "comments notes"}:
+        return ""
+    if "true information" in v or v in {"true", "real"}:
+        return "true_information"
+    if "misinformation" in v or v in {"false", "fake"}:
+        return "misinformation"
+    if "not sure" in v or "unsure" in v:
+        return "not_sure"
+    return clean_text(value)
+
+
+def standardize_mist_judgment(value: object) -> str:
+    v = norm(value)
+    if not v:
+        return ""
+    if v in {"real", "true", "1"} or "real" in v:
+        return "true_information"
+    if v in {"fake", "false", "0"} or "fake" in v:
+        return "misinformation"
+    return clean_text(value)
+
+
+def age_bucket(value: object) -> str:
+    s = clean_text(value)
+    if not s:
+        return ""
+    m = re.search(r"\d+", s)
+    if not m:
+        return s
+    age = int(m.group())
+    if age < 30:
+        return "18-29"
+    if age < 45:
+        return "30-44"
+    if age < 60:
+        return "45-59"
+    return "60+"
+
+
+def education_bucket(value: object) -> str:
+    v = norm(value)
+    if not v:
+        return ""
+    completed = [
+        "bachelor",
+        "master",
+        "doctor",
+        "professional",
+        "college degree",
+        "university",
+        "graduate",
+    ]
+    if any(token in v for token in completed):
+        return "completed"
+    return "not_completed"
+
+
+def load_pandora() -> tuple:
+    if not PANDORA_PATH.exists():
+        return [], [], []
+
+    raw = pd.read_csv(PANDORA_PATH, header=None, dtype=str, low_memory=False)
+    qids = [clean_text(x) for x in raw.iloc[0].tolist()]
+    labels = [clean_text(x) for x in raw.iloc[1].tolist()]
+    columns = []
+    seen: dict = {}
+    qid_map: dict = {}
+    for qid, label in zip(qids, labels):
+        base = label or qid or "unnamed"
+        seen[base] = seen.get(base, 0) + 1
+        col = base if seen[base] == 1 else f"{base}__{seen[base]}"
+        columns.append(col)
+        if qid:
+            qid_map[col] = qid
+
+    # Qualtrics export rows: machine ids, question text, choice/import rows, then responses.
+    df = raw.iloc[4:].copy()
+    df.columns = columns
+    news_cols = [c for c in df.columns if norm(c).startswith("news")]
+
+    claims: dict[str, dict] = {}
+    judgments: list = []
+    instances: list = []
+
+    # In this Qualtrics export the answer to "Your responses" is stored in the
+    # same column as the long claim prompt; the next two columns are heard-before
+    # and free-text notes, which are intentionally not released.
+    response_cols = {c: c for c in news_cols}
+    demo_cols = {
+        "gender": ["What is your gender?", "gender"],
+        "age_group": ["How old are you?", "age"],
+        "living_area": ["How would you describe the area you live in?", "area", "living_area"],
+        "education_bucket": [
+            "What is the highest level of education you have completed?",
+            "degree",
+            "education",
+        ],
+    }
+    id_candidates = ["ResponseId", "Response ID", "PROLIFIC_ID"]
+
+    for row_idx, row in df.iterrows():
+        row_judgments = []
+        participant_id = anonymize("pandora", row_idx, first_nonempty(row, id_candidates))
+        demographics = {}
+        for key, candidates in demo_cols.items():
+            val = first_nonempty(row, candidates)
+            demographics[key] = education_bucket(val) if key == "education_bucket" else clean_text(val)
+
+        if demographics.get("age_group"):
+            demographics["age_group"] = age_bucket(demographics["age_group"])
+
+        for news_col in news_cols:
+            prompt_text = clean_text(row.get(news_col, ""))
+            resp_col = response_cols.get(news_col, "")
+            judgment = standardize_pandora_judgment(row.get(resp_col, "")) if resp_col else ""
+            claim_text = extract_claim(prompt_text)
+            if not claim_text or not judgment:
+                continue
+            claim_id = qid_map.get(news_col, "") or anonymize("pandora_claim", news_col)
+            claims[claim_id] = {
+                "source": "PANDORA",
+                "claim_id": claim_id,
+                "claim_text": claim_text,
+                "gold_label": "",
+                "content_warning": True,
+            }
+            record = {
+                "source": "PANDORA",
+                "participant_id": participant_id,
+                "claim_id": claim_id,
+                "judgment": judgment,
+                **demographics,
+            }
+            judgments.append(record)
+            row_judgments.append(record)
+
+        for i in range(2, len(row_judgments)):
+            b1, b2, target = row_judgments[i - 2], row_judgments[i - 1], row_judgments[i]
+            instances.append(
+                {
+                    "source": "PANDORA",
+                    "instance_id": anonymize("pandora_instance", participant_id, target["claim_id"], i),
+                    "participant_id": participant_id,
+                    "target_claim_id": target["claim_id"],
+                    "target_judgment": target["judgment"],
+                    "observed_claim_1_id": b1["claim_id"],
+                    "observed_judgment_1": b1["judgment"],
+                    "observed_claim_2_id": b2["claim_id"],
+                    "observed_judgment_2": b2["judgment"],
+                    **demographics,
+                }
+            )
+
+    return list(claims.values()), judgments, instances
+
+
+def read_xlsx_first_sheet(path):
+    """Read a simple XLSX first sheet using only the Python standard library."""
+    ns = {"a": "http://schemas.openxmlformats.org/spreadsheetml/2006/main"}
+    with zipfile.ZipFile(str(path)) as zf:
+        shared = []
+        if "xl/sharedStrings.xml" in zf.namelist():
+            root = ET.fromstring(zf.read("xl/sharedStrings.xml"))
+            for item in root.findall(".//a:si", ns):
+                shared.append("".join(t.text or "" for t in item.findall(".//a:t", ns)))
+
+        sheet_name = "xl/worksheets/sheet1.xml"
+        root = ET.fromstring(zf.read(sheet_name))
+        rows = []
+        for row in root.findall(".//a:sheetData/a:row", ns):
+            values = []
+            for cell in row.findall("a:c", ns):
+                cell_type = cell.attrib.get("t", "")
+                value_node = cell.find("a:v", ns)
+                inline_node = cell.find("a:is/a:t", ns)
+                value = ""
+                if inline_node is not None:
+                    value = inline_node.text or ""
+                elif value_node is not None:
+                    raw = value_node.text or ""
+                    value = shared[int(raw)] if cell_type == "s" and raw.isdigit() and int(raw) < len(shared) else raw
+                values.append(value)
+            rows.append(values)
+
+    if not rows:
+        return pd.DataFrame()
+    width = max(len(r) for r in rows)
+    rows = [r + [""] * (width - len(r)) for r in rows]
+    header = [clean_text(x) or "column_{}".format(i) for i, x in enumerate(rows[0])]
+    return pd.DataFrame(rows[1:], columns=header)
+
+
+def load_mist_item_db() -> pd.DataFrame:
+    try:
+        db = pd.read_excel(MIST_ITEM_DB, dtype=str)
+    except ImportError:
+        db = read_xlsx_first_sheet(MIST_ITEM_DB)
+    cols = {norm(c): c for c in db.columns}
+    id_col = cols.get("id") or next((c for c in db.columns if norm(c) == "id"), "")
+    headline_col = (
+        cols.get("headline")
+        or cols.get("item")
+        or next((c for c in db.columns if "headline" in norm(c) or "item" == norm(c)), "")
+    )
+    out = db[[id_col, headline_col]].copy()
+    out.columns = ["item_id", "claim_text"]
+    out["gold_label"] = out["item_id"].map(lambda x: "true_information" if clean_text(x).upper().startswith("R") else "misinformation")
+    out["claim_id"] = [f"MIST_{i + 1}" for i in range(len(out))]
+    return out
+
+
+def load_mist() -> tuple:
+    if not MIST_PATH.exists() or not MIST_ITEM_DB.exists():
+        return [], [], []
+
+    df = pd.read_csv(MIST_PATH, dtype=str, low_memory=False)
+    item_db = load_mist_item_db()
+    claims = [
+        {
+            "source": "MIST-1",
+            "claim_id": row.claim_id,
+            "claim_text": clean_text(row.claim_text),
+            "gold_label": row.gold_label,
+            "content_warning": True,
+        }
+        for row in item_db.itertuples(index=False)
+    ]
+    claim_ids = item_db["claim_id"].tolist()
+    mist_cols = [c for c in df.columns if re.fullmatch(r"MIST_\d+", clean_text(c))]
+    if not mist_cols:
+        mist_cols = [c for c in df.columns if clean_text(c) in claim_ids]
+
+    judgments: list = []
+    instances: list = []
+    gender_cols = [c for c in df.columns if norm(c) in {"gender", "sex"} or "gender" in norm(c)]
+    age_cols = [c for c in df.columns if norm(c) == "age" or "age" in norm(c)]
+    edu_cols = [c for c in df.columns if "education" in norm(c) or "degree" in norm(c)]
+
+    for row_idx, row in df.iterrows():
+        demographics = {
+            "gender": first_nonempty(row, gender_cols),
+            "age_group": age_bucket(first_nonempty(row, age_cols)),
+            "living_area": "",
+            "education_bucket": education_bucket(first_nonempty(row, edu_cols)),
+        }
+        participant_id = anonymize("mist", row_idx)
+        row_judgments = []
+        for col in mist_cols:
+            judgment = standardize_mist_judgment(row.get(col, ""))
+            if not judgment:
+                continue
+            rec = {
+                "source": "MIST-1",
+                "participant_id": participant_id,
+                "claim_id": clean_text(col),
+                "judgment": judgment,
+                **demographics,
+            }
+            judgments.append(rec)
+            row_judgments.append(rec)
+
+        for i in range(2, len(row_judgments)):
+            b1, b2, target = row_judgments[i - 2], row_judgments[i - 1], row_judgments[i]
+            instances.append(
+                {
+                    "source": "MIST-1",
+                    "instance_id": anonymize("mist_instance", participant_id, target["claim_id"], i),
+                    "participant_id": participant_id,
+                    "target_claim_id": target["claim_id"],
+                    "target_judgment": target["judgment"],
+                    "observed_claim_1_id": b1["claim_id"],
+                    "observed_judgment_1": b1["judgment"],
+                    "observed_claim_2_id": b2["claim_id"],
+                    "observed_judgment_2": b2["judgment"],
+                    **demographics,
+                }
+            )
+
+    return claims, judgments, instances
+
+
+def load_wvs_priors() -> list:
+    records: list = []
+    for axis, path in WVS_FILES.items():
+        if not path.exists():
+            continue
+        df = pd.read_csv(path, dtype=str)
+        q_col = next((c for c in df.columns if norm(c) in {"question", "question text"}), "")
+        id_col = next((c for c in df.columns if norm(c) in {"question no", "question_no", "qid"}), "")
+        for _, row in df.iterrows():
+            for col in df.columns:
+                n = norm(col)
+                if not n.endswith("distribution") or n == "overall distribution":
+                    continue
+                group = re.sub(r"_?distribution$", "", col).strip("_")
+                most_col = f"{group}_most"
+                least_col = f"{group}_least"
+                records.append(
+                    {
+                        "demographic_axis": axis,
+                        "group": group,
+                        "question_id": clean_text(row.get(id_col, "")) if id_col else "",
+                        "question_text": clean_text(row.get(q_col, "")) if q_col else "",
+                        "distribution": clean_text(row.get(col, "")),
+                        "most_common": clean_text(row.get(most_col, "")),
+                        "least_common": clean_text(row.get(least_col, "")),
+                        "source_file": path.name,
+                    }
+                )
+    return records
+
+
+def main() -> None:
+    OUT.mkdir(parents=True, exist_ok=True)
+    p_claims, p_judgments, p_instances = load_pandora()
+    m_claims, m_judgments, m_instances = load_mist()
+    wvs = load_wvs_priors()
+
+    claims = pd.DataFrame(p_claims + m_claims).drop_duplicates(["source", "claim_id"])
+    judgments = pd.DataFrame(p_judgments + m_judgments)
+    instances = pd.DataFrame(p_instances + m_instances)
+    priors = pd.DataFrame(wvs)
+
+    claims.to_csv(OUT / "claims.csv", index=False)
+    judgments.to_csv(OUT / "judgments.csv", index=False)
+    instances.to_csv(OUT / "evaluation_instances.csv", index=False)
+    priors.to_csv(OUT / "wvs_group_priors.csv", index=False)
+
+    summary = {
+        "claims": len(claims),
+        "judgments": len(judgments),
+        "evaluation_instances": len(instances),
+        "wvs_group_prior_rows": len(priors),
+        "sources": {
+            "PANDORA": {
+                "claims": int((claims["source"] == "PANDORA").sum()) if not claims.empty else 0,
+                "judgments": int((judgments["source"] == "PANDORA").sum()) if not judgments.empty else 0,
+                "evaluation_instances": int((instances["source"] == "PANDORA").sum()) if not instances.empty else 0,
+            },
+            "MIST-1": {
+                "claims": int((claims["source"] == "MIST-1").sum()) if not claims.empty else 0,
+                "judgments": int((judgments["source"] == "MIST-1").sum()) if not judgments.empty else 0,
+                "evaluation_instances": int((instances["source"] == "MIST-1").sum()) if not instances.empty else 0,
+            },
+        },
+        "privacy": {
+            "dropped": [
+                "ResponseId",
+                "PROLIFIC_ID",
+                "IPAddress",
+                "LocationLatitude",
+                "LocationLongitude",
+                "open-text notes/comments",
+                "raw timestamps",
+            ],
+            "participant_ids": "deterministic salted SHA-256 hashes, truncated to 16 hex characters",
+            "demographics": "broad gender, age bucket, education bucket, and living-area fields only",
+        },
+    }
+    (OUT / "dataset_summary.json").write_text(json.dumps(summary, indent=2))
+    print(json.dumps(summary, indent=2))
+
+
+if __name__ == "__main__":
+    main()