Hi there, Saif!

First of all, if you get the answer to the 1st question, please , let me know! : )

Secondly, I would like to know why there is such a recommendation for discarding 'bot posts'. Personally, If I were able to identify these posts, I'd be curious to know WHAT they repeat (e.g. in terms of topic) , WHEN it occurs, WHAT KIND OF ASSOCIATIONS these repetitions generate when THE TEXT READS THE READER.

So, I guess the data can be rather precious to trace alternative connexions and relations , depending upon the way we look at them.

Nice topic!

All the best

It is a concern because a single bot can influence public opinion, for example, by sending posts with hate speech. But frankly speaking, I also wonder why somebody trusts posts in the social media if everybody knows that such posts can be toxic. Isn't it easier not to read and not to trust such posts?

Echoing Alexander the influence of bots on results of online polls, feedback forms, and Twitter has been noted in many political conversations including the 2016 US Election and the FCC Feedback Form on Net Neutrality provisions.

With the FCC case more than 22 million comments were received. Of these, 57% were determined to be from temporary or duplicate email addresses, seven comments represented 38% of all submissions, and only 6% were found to be unique. While it is possible to see this type of skew when a large political organization uses a form-letter via a website, the sheer number of comments on a relatively obscure provision (prior to commentator John Oliver's reporting) was very surprising. Pew Research Center did an analysis of the comments to measure the scale of the problem: http://www.pewinternet.org/2017/11/29/public-comments-to-the-federal-communications-commission-about-net-neutrality-contain-many-inaccuracies-and-duplicates/

Unfortunately, there isn't a defined way to identify bots, and they will become more sophisticated every time we find a way to identify them. An article from BuzzFeed points out some common ways to identify bots, many of which can be operationalized: Short-lived accounts, "bursting" of posts, concentration on only a limited number of news outlets, etc. (https://www.buzzfeed.com/lamvo/twitter-bots-v-human). Additionally, the techniques used above by Pew Research Center such as looking for similar or exact duplication of text.

Thanks, Philip. Saif, I would identify bot texts using NLP. As far as I know, there is no really mature technology to generate texts, in your case, post texts. I was thinking about how I would still approach the task to generate texts and to imitate conscious intelligent post texts in a thread in the social media. The easiest way to do it is to construct patterns for post texts and to identify terms that are discussed in the thread. After the identification, the task is to substitute particular terms to imitate an intelligent offensive reply. For example, if I detect that a post is about Trump, I can use the pattern "How do you know that Obama is better than ?" and resolve the pattern as "How do you know that Obama is better than Trump?" Hence, if I identify users that use particular patterns too frequently, I can suspect them to be bots and my task would be identification of such patterns. Is it clear what I mean?

Best, Alexander

Thank you, Jaime, Alexander, and Philip. These are very helpful responses and I am glad my questions provoked a discussion about the solution as well as the larger issue. My second question was honestly rhetorical. I know that the 2016 election generated a lot of concern about bots and reviewers, too, are therefore anxious about how they might affect analyses of social data, especially about political issues. I just find it ridiculous that bots (or any tech) are viewed as a problem when they undermine the American political process.

Before that, when the same or similar technologies were used by Americans/Europeans to disrupt politics around the world, they are viewed as a boon for democracy. Even now, I don't think anyone will question the validity of all those 100,000 studies that claim Twitter and Facebook enabled the Arab Spring and protests in Latin America etc. No one will ask whether those revolutions were also driven by (American/Western) bots — just as no one questions the role of American/Western governments in destabilizing Iraq, Syria, Libya, and so on while blaming everyone else for the violence. It's amazing how scholars, supposedly the vanguard of critical and independent thinking in a society, are driven by the same ideological impulses that drive governments.

Saif, in relation to the disruptions you've mentioned, I'd see them, from a perspective of Discourse, as a (novel) kind of selective "rationalisation"!

The emergent relation between Digitalisation and Ideology can produce such frames of dispute and selective repurposing ( for instance, affecting what we understand as History). So, are bots really to be discarded? I'd keep them to produce alternative points of view...

Philip, thank you for the links!

All the best