Back to the MIT repository
4. Malicious Actors & Misuse2 - Post-deployment

Society Manipulation

manipulation of social dynamics

Source: MIT AI Risk Repositorymit470

ENTITY

2 - AI

INTENT

1 - Intentional

TIMING

2 - Post-deployment

Risk ID

mit470

Domain lineage

4. Malicious Actors & Misuse

223 mapped risks

4.1 > Disinformation, surveillance, and influence at scale

Mitigation strategy

1. Mandate the implementation of a comprehensive, risk-based framework for **Algorithmic Auditing and Transparency**. This framework must require regular fairness audits, explainability reports, and the objective verification of data inclusivity across linguistic, demographic, and regional variables to systematically prevent the upholding of current biases, ensure system integrity, and promote the accountability of content-curating algorithms. 2. Enforce the **Mitigation of Information Monocultures and Filter Bubbles** through engineering controls. This necessitates the design and deployment of recommendation system mechanisms that proactively integrate viewpoint diversity, exposing users to a plurality of perspectives and content that challenges existing beliefs, rather than exclusively reinforcing them via past engagement data. 3. Establish and resource national-level programs for **Digital and Media Literacy Education** to build cognitive resilience. These programs must be systematically integrated to cultivate critical thinking skills, enhance the ability of individuals to recognize manipulative digital strategies, and empower users to critically evaluate the provenance and veracity of AI-generated or algorithmically amplified information.

ADDITIONAL EVIDENCE

allegations have been made against social media platforms for employing algorithms to control the content that users encounter in their feeds, with the aim of advancing specific political or commercial agendas. Personalized search algorithms strive to enhance the search process for users by presenting results that align with their interests and requirements, aiming for a more efficient experience. Nevertheless, these algorithms have faced criticism for their potential to uphold current biases, restrict information diversity, and foster filter bubbles