VL:AI

Personality computing is the use of artificial intelligence and big data to infer stable psychological traits from observable data sources.

• These data include behavioral residues, digital footprints, visual cues, and language use.

• The goal is to reconstruct or predict personality traits without relying exclusively on self-report questionnaires.

1. Identity claims are explicit statements about oneself or others (e.g., profiles, self-descriptions).

2. Behavioral residues are indirect traces left by behavior or appearance (e.g., likes, posting frequency, facial expressions).

   
   

   • Personality computing primarily relies on behavioral residues, which are often less consciously controlled.

Digital footprints are behavioral residues, appearance cues, and online traces used to make predictions about psychological traits.

1. Common sources include:

   • Social media data (likes, posts, images, videos)

   • Textual data analyzed with Natural Language Processing (NLP)

   • Visual data analyzed with machine vision

   • Smartphone and sensor data (location, app use, communication patterns) Combining multiple sources often improves prediction accuracy.

• starts with big data collection,

• followed by data preprocessing, feature extraction, and AI model training and testing.

• The final output consists of predictions or classifications of psychological traits, usually validated against self-report measures.

1. Personality Recognition – ground truth: self-report

2. Personality Perception – first impressions: peer reports

3. Personality Synthesis – human–machine interface

• Identifying trait-relevant content and how traits manifest online

• Identifying group differences (e.g., gender, age, culture)

Personality traits can be accurately predicted from social media likes, often outperforming human judgments.

1. Data acquisition (e.g., web scraping)

2. Data preprocessing

3. Textual feature extraction and selection

4. Classification or regression

5. Assignment of personality labels

1. Closed vocabulary: based on predefined linguistic categories and theory

2. Open vocabulary: discovers new linguistic features from data

1. 12 studies (2022)

2. Cross-cultural samples (US, Japan, Turkey)

3. Personality measures: Big Five, MBTI

4. Methods: SVM, Naive Bayes, Random Forest, Deep Learning

5. Average accuracy ≈ 78% (range 59–98%)

• A method to classify emotional content in text,

• with reported accuracies between 60–95% across domains such as cyberhate, social, and political content.

To predict personality traits from visual cues,

• such as facial expressions, movements, and posts, often using multimodal data.

1. Personality perception: first impressions (peer reports)

2. Personality recognition: real personality (self-reports)

• App usage, (Openness,Conscientiousness)

• communication patterns, (Conscientiousness, and Extraversion.)

• day–night use, (Conscientiousness, and Extraversion.)

• and music preferences (Openness)

• can predict traits such as ,

• Personality synthesis refers to designing artificial systems (e.g., chatbots, social robots)

• so that users perceive them as having stable, well-defined personality traits, based on artificial cues generated by the system.

When artificial agents become highly human-like but not perfectly realistic, users may experience discomfort or distrust.

This limits how far human-likeness should be pushed in social robots or avatars.

rule-based systems

, machine-learning mappings,

and neural or generative models that control language, behavior, or appearance to convey personality traits.

• Rule-based systems generate personality-related behavior using explicit, predefined rules created by designers.

• Personality traits are mapped to if–then rules that control language, behavior, or appearance.

• These systems are transparent and easy to interpret but lack flexibility and adaptability.

• Machine learning mapping uses data-driven models to learn associations between personality traits and observable cues (e.g., language patterns, gestures). I

• nstead of predefined rules, the system learns mappings automatically from training data.

• Neural models use artificial neural networks to model complex, non-linear relationships between personality traits and behavior.

• They can generate nuanced responses and adapt to context but are often black-box systems.

• Generative models create new content (e.g., text, speech, gestures) rather than selecting from predefined options.

• They can express personality dynamically and coherently, especially in language-based systems.

• Conversational agents are AI systems designed to interact with users via natural language.

• Personality synthesis is used to make these agents appear consistent, engaging, and socially appropriate.

user trust, engagement, and perceived intelligence, and helps users form stable expectations about the agent’s behavior.

• interact with humans in social contexts using speech, facial expressions, gestures, and movement. Personality synthesis supports natural and intuitive interaction.

• non-verbal behavior, emotional expression, communication style, and interaction strategies, helping humans interpret and predict robot behavior.

conversational agents designed to support mental health interventions, such as emotional support, psychoeducation, or therapy assistance. Personality is used to promote empathy, warmth, and consistency.

• quick, automatic

• data driven approach

• contributions are bidirectional

• questionnaires are goldstandard, but measurement error is not adressed (Big5 vs. mbti vs Hexaco)

• visibility of some traits

• ethical issues