[Paper Review] Beyond Personalization: Research Directions in Multistakeholder Recommendation

Himan Abdollahpouri, Gediminas Adomavičius+6 more|arXiv (Cornell University)|May 1, 2019

Recommender Systems and TechniquesComputer Science151 references66 citations

TL;DR

The paper articulates a multistakeholder framework for recommender systems, outlining stakeholder types, a three-dimensional design space, and research directions across economics, multi-objective optimization, and reciprocal/matching personalization.

ABSTRACT

Recommender systems are personalized information access applications; they are ubiquitous in today's online environment, and effective at finding items that meet user needs and tastes. As the reach of recommender systems has extended, it has become apparent that the single-minded focus on the user common to academic research has obscured other important aspects of recommendation outcomes. Properties such as fairness, balance, profitability, and reciprocity are not captured by typical metrics for recommender system evaluation. The concept of multistakeholder recommendation has emerged as a unifying framework for describing and understanding recommendation settings where the end user is not the sole focus. This article describes the origins of multistakeholder recommendation, and the landscape of system designs. It provides illustrative examples of current research, as well as outlining open questions and research directions for the field.

Motivation & Objective

• Motivate the shift from pure user-centric personalization to incorporating multiple stakeholders in recommendation outcomes.
• Introduce formal definitions and a three-dimensional design space for multistakeholder recommendations (Consumers, Providers, System).
• Survey existing research areas (economics of multisided platforms, multi-objective recommendation, and matching/personalization) and outline open questions and directions.

Proposed method

• Define a formal recommender S as a function f mapping (u, i, R) to a score, and distinguish user-oriented versus multistakeholder outputs.
• Introduce stakeholder triplet notation for configurations: Consumers (C), Providers (P), System (S) with active/passive and personalized/neutral variants.
• Present a three-dimensional design landscape for <C, P, S> including S variants such as neutral, aggregate, and targeted objectives.
• Classify multistakeholder applications with a taxonomy (e.g., Table 1) showing configurations and associated design choices.
• Discuss connections to economics (multisided platforms), multi-objective optimization, and reciprocal/matching personalization.
• Describe illustrative examples and domains, including reciprocal recommendation and system-oriented fairness.

Experimental results

Research questions

• RQ1 What are the essential stakeholder roles and interaction modes in multistakeholder recommendation?
• RQ2 How can we model and classify multistakeholder recommendation designs across consumer, provider, and system perspectives?
• RQ3 What research directions and open questions arise from applying multisided platform theory and multi-objective optimization to recommender systems?
• RQ4 How do reciprocity and fairness concepts integrate into multistakeholder frameworks?

Key findings

• A formal multisided framework extends traditional recommender design to include Consumers, Providers, and System as three interacting stakeholders.
• A three-dimensional <C, P, S> design space captures active/passive interactions and neutral/personalized personalization across stakeholders.
• Economics of multisided platforms and fairness/diversity/objective balancing motivate multistakeholder approaches beyond user-centered metrics.
• Reciprocation and matching concepts (e.g., reciprocal recommendation) are central in several domains like online dating and job matching.
• The paper provides illustrative examples and identifies largely unexplored areas, especially S_t (system-targeted) designs, as research directions.
• The framework highlights the gap between academic user-centric research and real-world commercial objectives that integrate multiple stakeholders.