The SHREC Wellness Home subset contains real-world, longitudinal interaction video data from Jeong et al. (2023) recordings from an 8-week in-home study with adult participants aged 18–83. Participants engaged with a socially assistive robot designed to improve psychological well-being, affect, and readiness for change through evidence-based positive psychology interventions (PPIs).

Authors: Dong Won Lee, Yubin Kim, Sooyeon Jeong, Denison Guvenoz, Parker Malachowsky, Louis-Philippe Morency, Cynthia Breazeal, Hae Won Park
Institutions: MIT, Purdue University, Carnegie Mellon University
Code Repository: https://github.com/mitmedialab/SHREC

‼️ Notice

To request access to the SHRECt, please apply using your official institutional affiliation email address. Requests submitted from personal email accounts may be delayed or rejected, as access is intended for verified research and academic use.

🧠 SHREC Dataset Summary

Subset Info: Subset of SHREC Dataset (Original Raw Video Data From A Robotic Companion for Psychological Well-being: A Long-term Investigation of Companionship and Therapeutic Alliance)

~78 real-world interaction videos
~1900 expert annotations

In full, the Social Human Robot Embodied Conversation (SHREC) Dataset is the first large-scale, real-world benchmark designed to evaluate social reasoning in language and vision-language models through physically embodied human-robot interactions (HRI). It contains:

~400 real-world interaction videos
~10K+ trained human annotations
Labels for social errors, competencies, rationales, and corrections
Coverage of seven social attributes critical for social intelligence

It supports research in rapport-building, mental health intervention, and social reasoning in intimate, longitudinal HRI settings.

💾 Load Dataset

To load this dataset into a pandas df:

import pandas as pd
from datasets import load_dataset
import pandas as pd
import glob
import os
from huggingface_hub import snapshot_download

repo_id = "MIT-personal-robots/shrec_wellness_home"

snapshot_download(repo_id=repo_id, repo_type="dataset", local_dir="shrec_wellness_home", token = True)

parquet_files = glob.glob("downloaded_repo/**/*.parquet", recursive=True)

df = pd.concat([pd.read_parquet(f) for f in parquet_files], ignore_index=True)

📦 Dataset Structure

Each interaction sample includes:

video_id: Identifier for the interaction session
frame_paths: List of image paths (15 selected frames from the video)
video_path: Path to the anonymized interaction video file ({video_id}/{video_id}.mp4)
transcript: Multi-turn dialogue between user and robot
label: "competence", "error", or "none"
social_attributes: List of relevant attributes from 7 core categories
rationale: Explanation for the error or competence
correction: Suggested repair if the segment is an error

🧪 Benchmark Tasks

The HSRI benchmark includes 8 tasks spanning four core dimensions of social reasoning in human-robot interaction.

1. Detecting Social Behavior

Task	Description
Error / Competence / None Detection	Classify the robot’s behavior as a social error, competence, or neither.
Error Detection	Determine whether a given behavior constitutes a social error.

2. Identifying Social Attributes

Task	Description
Social Attribute Identification	Identify which of the seven social attributes are relevant to a given behavior.
Multiple Attribute Detection	Determine whether multiple social attributes are present in the behavior.

7 Social Attributes

Emotions – Identifying and responding to emotional expressions
Engagement – Monitoring user interest and presence
Conversational Mechanics – Managing turn-taking, timing, and pauses
Knowledge State – Tracking shared knowledge and references
Intention – Inferring the goals or motives behind actions
Social Context & Relationships – Acting appropriately based on context and social role
Social Norms & Routines – Following culturally appropriate social conventions

3. Understanding Interaction Flow

Task	Description
Pre-Condition Reasoning	Given the robot’s utterance, choose the plausible user behavior that came before.
Post-Condition Reasoning	Given the user’s utterance, select the robot’s likely follow-up behavior.

These tasks are structured as multiple-choice questions, with distractors sampled from other real-world robot-user interactions to ensure contextual relevance.

4. Rationalizing & Correcting Social Errors

Task	Description
Rationale Selection	Choose the correct explanation for why the robot’s behavior was an error.
Correction Suggestion	Select the most appropriate corrective action the robot should have taken instead.

These tasks test a model's diagnostic (understanding what went wrong) and prescriptive (knowing how to fix it) social reasoning abilities.

🔍 Example Sample

{
  "ID": "P15_s002-006",
  "sample_frame": "P15_s002-006/0000.png",
  "transcript": "AI Agent: (00:00:02) Hey there. How was your day today?\nUser A: (00:00:04) Good. How was yours?\n...\nAI Agent: (00:10:42) ... brighten our days.",
  "Annotations_A": [
    {
      "timestamp": {"start": 7.21, "end": 20.23},
      "error": true,
      "source": {"Verbal": true, "Non-Verbal": false},
      "attribute": {
        "Conversational Mechanics": true,
        "Intention": false,
        "Emotions": false,
        "Engagement": false,
        "Knowledge State": false,
        "Social Context &  Relationships",
        "Social Norms & Routines"
      },
      "rationale": "Delayed response and failure to understand participant.",
      "correction": "Should have responded within 2–3 seconds."
    }
  ],
  "Annotations_B": [
    {
      "..."
    }
  ],
  "Annotations_C": [
    {
      "..."
    }
  ],
  "framerate": 15.0,
  "frame_paths": [
    "P15_s002-006/0000.png",
    "P15_s002-006/0013.png",
    "P15_s002-006/0034.png",
    "..."
  ]
}

Downloads last month: 99

Total file size:

32.7 GB

Collection including MIT-personal-robots/shrec_wellness_home

SHREC

Collection

Social Human Robot Embodied Conversation Dataset: a multimodal video dataset of human robot interaction to benchmark AI agents' social Intelligence. • 3 items • Updated May 15, 2025 • 4