Twitter’s new Responsible ML initiative aims to address algorithm challenges

Twitter said it is conducting in-depth analysis and studies to assess the existence of potential harms in its algorithms.

Twitter launches a new initiative called Responsible ML

Twitter has introduced a new initiative under which it is trying to improve its Machine Learning (ML) algorithms and ensure they do not cause unintentional harms.

Twitter also has an ML Ethics, Transparency and Accountability (META) team that consists of a dedicated group of researchers, data scientists, and engineers looking into these ML-related challenges. The team will research the following aspects on priority:

Gender and racial bias analysis of image cropping (saliency)

Assessment of Home timeline recommendations across racial subgroups

Analysis of content recommendations for different political ideologies

“We're also building explainable ML solutions so you can better understand our algorithms, what informs them, and how they impact what you see on Twitter. Similarly, the algorithmic choice will allow people to have more input and control in shaping what they want Twitter to be for them. We're currently in the early stages of exploring this and will share more soon,” said Twitter in a blog post.

Excerpts from Twitter's new blog on Responsible ML (Twitter)

Excerpts from Twitter's new blog on Responsible ML (Twitter)

Twitter has introduced a new initiative under which it is trying to improve its Machine Learning (ML) algorithms and ensure they do not cause unintentional harms.

Twitter also has an ML Ethics, Transparency and Accountability (META) team that consists of a dedicated group of researchers, data scientists, and engineers looking into these ML-related challenges. The team will research the following aspects on priority:

Gender and racial bias analysis of image cropping (saliency)

Assessment of Home timeline recommendations across racial subgroups

Analysis of content recommendations for different political ideologies

“We're also building explainable ML solutions so you can better understand our algorithms, what informs them, and how they impact what you see on Twitter. Similarly, the algorithmic choice will allow people to have more input and control in shaping what they want Twitter to be for them. We're currently in the early stages of exploring this and will share more soon,” said Twitter in a blog post.|#+|

Twitter is also asking the public in helping shape up its ResponsibleML initiative. Interested users can also learn more about the programme, using #AskTwitterMETA on Twitter.

Machine learning may seem technical but it's what helps us show you relevant Tweets and Topics, and take action on Tweets that may break our rules.

We're sharing more about the work we're doing to develop responsible ways to use ML to benefit you: https://t.co/FOFYH36TCe

— Twitter Support (@TwitterSupport) April 14, 2021

The latest announcement comes as part of Twitter's efforts to improve its algorithms to address widely reported problems such as bias, harassment, and the spread of misinformation. The company has drawn flak for not doing enough to curb these.

In the case of photo crop, Twitter acknowledged its automated image cropping feature had an issue wherein the algorithms consistently picked up and previewed lighter-skinned people regardless of the framing of the original image.

Apart from fixing the algorithms, Twitter has also sought public input on devising a framework for content posted by politicians and government officials. Apart from conducting the survey, Twitter is also taking help from human rights experts, civil society organisations and academics globally.