Crowdsourced ethics with personalized story matching

Abstract

Cyberbullying is a widespread and growing social problem, threatening the viability of social networks for youth. We believe that one of the best ways to combat this problem is to use these incidents as "teaching moments", encouraging teens to reflect on their behavior and choices. Sites that offer community discussions around the ethical aspects of social situations can help teens feel less alone in their plight, and provide useful advice and emotional support. But the success of these "crowdsourced ethics" sites depends critically on the user feeling like discussions are relevant to their own personal experience.We have augmented the crowdsourced ethics site "Over The Line", offered by MTV Networks, with a personalized story matcher that classifies stories according to dynamically discovered high-level themes like "sending nude pictures online" or "feeling pressure in relationships". The matcher uses a mixed-initiative LDA machine learning technique [2], and a commonsense knowledge base specialized to the bullying problem. The site is currently public, and attracts an audience of thousands of users daily.

Online Concerns of Parents Suspecting Autism Spectrum Disorder in Their Child: Content Analysis of Signs and Automated Prediction of Risk

“…In another study, adolescents at risk of being bullied were identified using online texts from MTV’s A Thin Line project. Thousands of teenagers’ online posts were used to develop an algorithm for identifying offensive cyber bullying, based on topic modeling methods [34]. This study built upon these works to test the possibility of devising an automated ASD-risk estimator online, based on large amounts of unstructured textual data and a machine learning algorithm.…”

Achieving Ethical Algorithmic Behaviour in the Internet of Things: A Review

“…What is proposed is for tools to be developed that can take technical aspects of algorithms and present them to the general public so that they can be engaged in influencing the effect and behaviour of the algorithms, effectively crowdsourcing ethics, an approach used elsewhere [101]. The general approach of combining machine-learned representations and human perspectives is also called lensing (https://www.media.mit.edu/videos/2017-05-18-karthik-dinakar/, accessed on 1 July 2021).…”

Hair

“…For example in the field of machine learning, crowd workers have been employed to generate ground truth data and features and to validate results after classification [7,9]. Alternatively, other systems have utilized machine learning and classification techniques to filter low quality crowdsourced data [25,42,45]. However, it is not until recently that crowdsourcing has been combined with machine computations to improve accuracy of tasks that are difficult for either to perform independently [10].…”