New day, new writing task. This is a reflection I wrote for the Social Computing course. I will tackle this in three parts, based on the references linked below:
The first topic for this analysis is “Crowdsourced Detection of Emotionally Manipulative Language”, as I have a particular interest on emotion understanding and human behavior, and goes on to distinguish emotional manipulative language (EML, language that induces a particular emotion in the reading to get a result on their benefit) from intrinsically emotional content (IEC, or information that is emotional regardless of the language used). There is a need to fix non-expert annotations because they often mix IEC with EML and it gets labeled wrong, generating false positives around half of the time. If machines were trained on this dataset, the error would propagate. To overcome this, the anchor comparison was a brilliant approach. This text has really amplified my view on moderation bots and the criteria that is used on forums like Reddit. If applications / forums start to include a “highly manipulative” label, people will be less prone to manipulation and coordinated attacks / harassment will be less likely to happen. The fact that this error can propagate on chatbots if not treated properly highlights the need for quality dataset for LLM models, since they are available to all public nowadays.
The second topic is “The nightmare videos of children's YouTube”. I was familiar with this topic of discussion beforehand, but it still amazes me how auto generated content can be so non language friendly (like “adds” with a language syntax that is not Grammarly correct but SEO friendly), and how it will impact kids in the future. Moreover, this is a huge wakeup call on fake news. Misinformation is spreading through small children, and if left unsupervised they can find overtly sexual videos or content that might generate phobias. Nowadays it is not enough to know one scope of software engineering, future professionals must consider ethical concerns like this one and use critical thinking to avoid fake videos like this to propagate. ?
The third topic chosen for this analysis is “Will the Crowd Game the Algorithm?”, which is relevant nowadays due to the increase of fake news and misinformation in social media (in particular, platforms like Facebook that is mainly used by middle-aged people). I agree that it was wise to use the “wisdom of the crowd” to effectively identify fake news website and keep thrusted content afloat. It is surprising to know that there was no bias on discernment in the group that was aware that their decisions will influence the social media algorithms score, as people with power usually abuse it. Of course, there was some minor struggle with political sources (given that as the text said, some people cannot discern between “fact” and “opinion”), but overall, the raking offered by the crowd often emulated those provided by professional fact-checkers.