Personality Aware Recommendation and Persuasion

62 concepts48 papers

Conventional recommendation systems excel at identifying what a user might like by analyzing their past interactions. If a user watches three sci-fi movies, the system recommends a fourth. However, this approach hits a hard ceiling when historical data is sparse, or when the objective shifts from passive consumption to active behavior change—such as convincing a user to donate to a charity, adopt healthier habits, or heed a security warning.

To break through this ceiling, a growing body of research focuses on psychometric personalization. Instead of treating users as mere vectors of item preferences, these systems model the underlying psychological traits—such as the Big Five personality dimensions—that govern how individuals process information. The core insight is that persuasiveness is a relational property, not an inherent quality of a message. By aligning the style, framing, and structure of an interaction with a user's psychological profile, systems can dramatically increase user satisfaction and intervention efficacy.


Persuasion as a Relational Property

In traditional information retrieval, an argument or description is often treated as having an objective "strength." Psychometric personalization rejects this assumption. As Berkovsky et al. argue in their work on susceptibility-based strategy adaptation, individuals possess highly varied sensitivities to different psychological influence mechanisms, such as social proof, authority, or scarcity.

This variance explains why a one-size-fits-all approach to digital interventions frequently fails. For example, in persuasive dialogue systems designed for charitable giving, a generic macro-argument might alienate a user if the rhetorical mode clashes with their cognitive style. Anand et al. demonstrated this audience effect empirically: while factual, structured monologues are generally robust, individuals who score high in Openness and Conscientiousness are significantly more receptive to emotional, affect-based arguments.

Similarly, El Baff et al. found that a reader's personality traits dictate whether they are persuaded by the substantive content of an editorial or its stylistic elements. For instance, individuals with high neuroticism and low extraversion exhibit a heightened susceptibility to the purely stylistic and linguistic aspects of a text.

[Standard Message] ───> [Universal Filter] ───> Variable Compliance

[Tailored Message] ──> [Matched Personality Filter] ──> High Compliance

Personality as a Domain-Independent Prior

One of the most compelling reasons why psychometric modeling works is that personality traits are stable, cross-situational characteristics. This makes them incredibly powerful as high-level psychological priors, particularly in data-sparse environments or cold-start scenarios where a user's domain-specific history is entirely missing.

Yang et al. leveraged this concept in their multi-modal recommender system, PersiC. By mining Myers-Briggs Type Indicator (MBTI) traits from social media behavior, the authors injected personality profiles directly into deep learning recommendation architectures. This psychological prior allowed the system to map user preferences accurately even when direct interaction data was minimal, yielding massive efficiency gains in real-world advertising deployment.

Furthermore, because these psychological profiles are universal, the framing logic learned in one domain can generalize to another. Ding et al. demonstrated this by mapping Big Five traits and core human values (via the Schwartz model) to content preferences. Their personalized emphasis framing model automatically selects which specific attributes of an item to highlight based on what resonates with the user's psychological profile, bypassing the need for extensive, domain-specific user histories.


The Mechanics of Tailoring

How does personality-aware adaptation manifest in system design? It operates across three primary dimensions: linguistic framing, structural composition, and explanation styles.

1. Linguistic Framing and Warning Systems

In high-stakes scenarios like mobile security, users routinely ignore technical warnings because the language fails to engage them. Wu et al. addressed this by automatically generating security-centric descriptions for applications tailored to individual personality traits. By shifting the linguistic focus to match the user's specific psychological concerns, the authors achieved a measurable increase in threat awareness and secure behavior compared to generic technical summaries.

2. Structural Composition of Content

Personality does not just dictate how a user reads a sentence; it dictates how they navigate choices. Ferwerda et al. studied how Big Five traits influence a user's preference for diversity within a recommendation list. They discovered that highly conscientious individuals are most satisfied when presented with high levels of variety, whereas highly agreeable users prefer moderate diversity. Tailoring the structural properties of the list itself prevents cognitive fatigue and increases satisfaction.

3. Explanation Styles

When an AI system explains why it recommended an item, the utility of that explanation depends heavily on the recipient's temperament. Using the REASONER dataset, Chen et al. established that personality traits significantly influence how users perceive different styles of explanations (e.g., text-based vs. video-based, or feature-level vs. user-level justifications). Incorporating these profiles allows systems to dynamically select explanation strategies that build trust rather than confusion.


Proactive and Conversational Adaptation

In interactive and conversational AI, psychometric awareness transforms how agents manage dialogue flow. Traditional proactive assistants often rely on rigid, universal interaction policies that assume a cooperative user. However, real humans frequently deviate from expected paths when they become bored or frustrated.

To handle these dynamics, systems must balance task-oriented goals with psychological comfort. The I-Pro policy developed by Lei et al. uses reinforcement learning to dynamically adjust dialogue strategies based on diverse psychological factors and behavioral signals.

This adaptive capability is critical for proactive agents that initiate contact. As Kim et al. demonstrated with their ProPerSim framework, a user's receptivity to proactive suggestions is highly contingent on their personality. For instance:

  • Users with low extraversion respond poorly to proactive prompts involving highly social environments, preferring recommendations for solitary activities.
  • Users with high openness are uniquely receptive to novel, creative, or unexpected prompts.

By simulating these interactions over extended periods, agents learn to evolve from generic recommenders into context-aware companions that respect individual psychological boundaries.

Therapeutic and Pedagogical Efficacy

The ultimate validation of this paradigm is found in behavioral change fields like digital health and education. Gjoreski et al. designed intelligent cognitive assistants that map Big Five traits to specific intervention delivery styles for stress and anxiety relief. The authors note that ignoring these stable traits drastically reduces therapeutic engagement.

This is echoed in physical deployments: when Jeong et al. deployed a robotic positive psychology coach in college dormitories, they found that the intervention's success was tightly coupled with the students' personality profiles—highly conscientious and less neurotic individuals showed the most significant well-being gains, proving that the robot's conversational strategy must be tailored to maximize therapeutic impact across diverse temperaments.

Key Foundational Work

Editorial for intelligent interactive multimedia systems and services (2012) — contributed 1 concept elements to this direction.
Argumentative Human Computer Dialogue for Automated Persuasion (2008) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_personality_based_recommendation_69657c3e,
  title = {personality-based recommendation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/personality-based-recommendation/rdr_69657c3e58b540f7a9549c4c1ccad16f},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_personality_based_recommendation_69657c3e_v1,
  title = {personality-based recommendation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/personality-based-recommendation/rdr_69657c3e58b540f7a9549c4c1ccad16f/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake