Dear All,
Today I wish to present you an article about automatic detection of political ideology using recursive neural networks. In recent days there were discussed problems of privacy and influence of social media on political views. In this article the authors use neural networks, which are very efficient in solving many problems, to identify political views based on spoken and written language features.
http://www.aclweb.org/anthology/P14-1105
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
2. Do you think that such tools are safe for our society and freedom of opinion?
3. Could it be used for good purposes?
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteWell probably other sources of political ideology, probably social media posts, maybe some speeches transcripts and many more. But for other topics than politics it can also be easily used. Anything that can be detected from data set that will classify into groups. Maybe psychological types, food preferences, orientation. It seems like it can be used for anything.
2. Do you think that such tools are safe for our society and freedom of opinion?
For me it is another psychological tool, that can be used either for sake of education and development or for annoying things like marketing and advertisement. Unfortunately it will be used in both cases not only good causes. Everyone by now should know that freedom of opinion basically does not exists and everything that we share publicly is tracked and used for some purposes.
3. Could it be used for good purposes?
As I said above for some psychological research and maybe some scientific work. It might be one of the steps for understanding human brain more. Of course I can only think up potential purposes, but is seems for me as a tool to achieve more understanding of human thinking.
Thank you for your comment and good examples of data of other topics than politics. I cannot fully agree that it can be used for better understanding of human thinking because neural networks usually work as a black box and we can only know the result.
Delete1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteI believe that the method described in the article can be used not only for the analysis of ideological views. Neural networks learn. If the source will be psychological profile books, you will be able to give an opinion about the state of human health.
2. Do you think that such tools are safe for our society and freedom of opinion?
No such tools are safe. We must be aware of what is safe on the net and what is not. As a society, we must learn to live in a virtual world without revealing so much information about ourselves.
3. Could it be used for good purposes?
Of course! everything can be used for good purposes. However, It is easier to poison someone's life - or make it difficult than to find "good use".
Thank you for your input. I agree that we must be aware what is safe what is not and that we are responsible for the informations we share.
DeleteAd1.
ReplyDeleteThe research was conducted in the USA. There is a two-party system. The representatives of these parties have opposing views. Therefore, these two parties usually compete with each other. For this reason, the arguments used in the debates have a high level of partisanship. Therefore, in this experiment binary classification was used. The result of the classification is the probability of determining the type of party (conservative or liberal) to which the data belong. Figure 4 shows that the deeper the neural network is learnt, the more data it is classified as neutral. The use of other data, e.g. social media, has no purpose, because such data have completely different input levels of partisanship. The results of such data will mostly be misclassified because they have been learned from other input data.
Ad.2
The results of such data may affect a group of people who are not sure which party they will vote for. Those who are undecided and more classified as conservatives may be persuaded to vote for the conservative party. Conversely, it does not.
Ad.3
It is not a good situation when some institution has information which may cause easier manipulation of another person.
Thanks for your comment. I agree that for USA the problem may be quite easier due to two-party system. Nevertheless, it is possible to make it multi classification problem and extend it on social media.
DeleteInteresting, but quite disturbing topic.
ReplyDelete1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
I wouldn't dismiss the Twitter idea, if I remember correctly it even had an API. Then there are blogs, which have RSS or Atom syndication, so it is quite machine-friendly. However, this is still likely to hit obstacles like Poe's law: https://en.wikipedia.org/wiki/Poe%27s_law
2. Do you think that such tools are safe for our society and freedom of opinion?
No, but not because of the machines itself, but because of people who are behind them and set them in motion. They have a convenient excuse "sorry, but it's not us -- our *algorithm* found you to be biased / not objective / fake news / etc.". To be continued under next question.
3. Could it be used for good purposes?
Like a benevolent totalitarian society? Well, it sure could be used with good intentions, but I don't think the overall purpose would be good. For now, this could help to sift through large volumes of data, and report suspected wrongthink to diligent human censors.
Other purposes could be also for example detecting heresy or witchcraft - which would be definitely tried, had Deep Learning and Recursive Neural Networks been developed at the times of the Inquisition.
If this technology became democratized, then it would also be a great tool for all egosurfers -- instead of manually monitoring the Internet for material about oneself, they could automatically get a notification on their smartphone if someone was writing badly about them on the Internet.
Thanks for your comment. Yes Twitter has its API and their data is widely analyses by scientists.
Delete1. Definitely, there are many possibilities regarding to such tools. I’ve been thinking about it since 2 days and now I believe it can even work for medical data. Look, everybody has some health problems, and everybody tried to look for some help in net. Either you are lost with detailed data, or you find bad news and focus on them, forgetting that each case is different and there are also some good news. Personally I used to find out something about my problems several years ago and now I know the medical data found in net can be very useful, but you have to use it accordingly. This is maybe the idea for recursive neural networks – because they bring possibility to gain more substantial, crucial information from the general huge set of data.
ReplyDelete2. In my opinion using such tools paradoxically could bring more safety in this respect, because they ensure more order in the data chaos we gain from net. this is my personal judgement, maybe polarized by my education, experience and trust directed towards technology – but my dream is to wake up in world with less, but better-quality information. I guess such tools can help to achieve such target in the future.
3. As a result of my assumptions made in point 1 and 2 above, I fully believe that it can be used for good purposes. Like all the tools, by the way, it depends on the humans, how they use the technology they’ve created.
Thank you for your comment. It is interesting idea to use this methods for medical data.
DeleteHi, it is interesting article.
ReplyDelete1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
I am not sure can we detect ideology when we are analyzing books or long text. In this kind of resources we can find examples that describes differences between ideologies and so on.
It could be interesting to analyze results of this neural network on politicians twitter accounts. Twitter has limited message length so we focus on one, short politician thought.
2. Do you think that such tools are safe for our society and freedom of opinion?
Generally speaking I am pretty sure that we are analyzed by thousands of neural network - on our social media. It allows corporations to sell us something or redirect us to the different webpage. Last days we have heard about political lobbying in social media. Our opinion is based in most of cases on that what we have read. That is power of social media.
3. Could it be used for good purposes?
Yes, for instance if we are able to detect terrorism, or person who has depression we can prevent disaster. I used word terrorism - so we can assume that a several AI bots will analyze context of this message.
Thanks for your input. I think similar approaches are used to detect terrorism.
DeleteAs authors are interested in a sentence level annotations the best match that comes to my mind is Twitter. Tweets contain only short statements that in general should be an essence of authors' opinion. We can assume that all tweets from a Liberal author represent Liberal opinions, so such dataset would be relatively easy to annotate. Other sources that I can think of would be blog posts.
ReplyDeleteI don't see any dangers in classifying political bias of a text. After all authors of such texts probably want the audience to know what political option they favor. The problem related to this is rather a censorship of media, but I think it happens without any sophisticated solutions. Automated filtering could only affect comments under online articles but those are usually trolling anyway, so many news sites just disable commenting feature.
It could be used for tagging online content for easier searching. Maybe it could even be a part of a larger service that would present all the popular news from both or more perspectives. Such presentation could lead to a less biased information.
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeletePublic TV appearances by politicians may also be used for this type of analysis. As well as records from political debates where we have to deal with people of extremely different views.
2. Do you think that such tools are safe for our society and freedom of opinion?
I believe that neural networks have much better applications. The analysis of political views will not help the public, but only politicians. This could lead to the situation that, using this knowledge and the recently mentioned manipulations in social media, the elections could be even more controlled by politicians.
3. Could it be used for good purposes?
It can help find people with extreme views that pose a danger to other citizens. However, I think that this would be too much cost for our privacy.
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteI think they can try to use classical books that made fundation of some doctrins for examples Marks and Engels books. But also data gather from social media should be fine. I am also thinking about taking political public speaches and made transriptions from them.
2. Do you think that such tools are safe for our society and freedom of opinion?
I think that in todays word you can technology in good and bad way. From one side we can use this tool to gain knowledge from gathered data but on the other side we can use it as a product placement or even it may be used by politicians in finding hooks for another one. So it all depends from the people.
3. Could it be used for good purposes?
Answer should be above this question :D.
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteI think that all the data we produce can show which ideology we are for. If this neural network works on our emails or posts in social media, the result will be the same.
2. Do you think that such tools are safe for our society and freedom of opinion?
I'm little confused about this problem. It's nothing suprise that TV program can create way of thinking. What haapen when algoritm start to do it better than all this TV programs together? I'm afraid of it, but we will see...
3. Could it be used for good purposes?
I think so. This neural network can potentially catch a dangerous person
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteAccording to me. a good source of data is any politically tagged statements. For example with Twitter. We have access to a diverse society in this way.
2. Do you think that such tools are safe for our society and freedom of opinion?
We can verify and investigate. But it should serve the public. Each type of information can be used for good and bad purposes. In this topic we are close to censorship.
3. Could it be used for good purposes?
I did not think about it more. But I imagine that researching the rhetoric of the debate can give interesting results. Maybe it will be possible to use this data in education ???
This article was very informative and useful for me, because politics is my hobby)
ReplyDelete1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
I think that the data from social networks could make this algorithm more informed. A lot of data from books on political topics. For example, Capital - Karl Marx or or 1984 Orwell. The newsline will also help to increase the database. Perhaps some political forums.
2. Do you think that such tools are safe for our society and freedom of opinion?
I think that it's not quite, because each system has its own errors.
3. Could it be used for good purposes?
Yes, of course, in most cases this system will be used for good purposes, if the algorithm of this system is error-free. But do not forget that he can never be perfect.some phrases the system will not be able to understand as subtly as a person, and can interpret with an error.
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteNeural networks are a very powerful tools. I think that the method will be useful for religious data.
2. Do you think that such tools are safe for our society and freedom of opinion?
I think it depends on who uses such tools and the domain of data. The tool can be helpful in suggesting decisions. Our freedom of speech will not be compromised because each of us has the right to express our opinion, believe and have our political views.
3. Could it be used for good purposes?
I think so, It depends on the human being. Each tool can be used for bad purposes as was the case with eg. P2P.
Thank you for a nice article. In my opinion, tools of this type can be used in general for the classification of different views and personality traits. The only problem is creating a good classification. I think such tools can be helpful in determining not only political views but also terrorist tendencies, mental illnesses, tastes and other characteristics that describe us. There may also be good applications such as detecting people who are dangerous to the environment or themselves... Unfortunately, I suppose that these kinds of tools will even more interfere with our privacy...
ReplyDelete1
ReplyDeleteI'd say the only way to a politicians mind is through direct connection to his/her neurons (if any available) not through words in any physical state as you might have figured out through life experience, words as containers can be quite deceptive. That's why Plato hated poets because they knew how to manipulate people just by playing with words. You'd rather want to analyse their actions instead - they can lead to a person's intention clearier.
2
No tool is safe or unsafe. A tool is a tool. Depends on the user. Example: imagine a metaverse where every idea has already been created. If you would be able to pick this tool, what would you use it for? It seems to work as an instrument to control someone (politician) who wants to control society (other people) so it's a mirror play, poetically speaking.
3
Back to the good/bad distinction. Cutting a tree might be good for you, eg. you might build a house. Although a tree is already someone's house so it depends how much you have to change what is to be satisfied :)
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeletePosts and user profiles data from social networks, normative acts and legislation documentation.
2. Do you think that such tools are safe for our society and freedom of opinion?
Tools itself are safe, but the way of usage determine if they do harm.
3. Could it be used for good purposes?
Simiraly as second question, yes it could be used for good puposes as well for the bad ones too. It is decided by the user.
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteI think about that such RNN framework could be used to determine in respect of which denomination the text was written. Maybe it could work also for ethical issues: does the text comply with ethical standards. I also consider verifying the text for compliance with historical facts.
2. Do you think that such tools are safe for our society and freedom of opinion?
I see few threats resulting from the use of such a solution. First of all, such a tool can be used to trace information, up to strict censorship, which is certainly a form of attack on human freedom. Secondly - manipulation of the texts of another party, e.g. in the pre-election period. Finally, the lack of the verification whether the operation of such a framework is impartial (I mean the selection of datasets for RNN analysis).
3. Could it be used for good purposes?
Certainly yes but maybe not in political ideology detection. If we talk about the use for politics, unfortunately I see only the negative usage. In the case of changing the datasets and area of analysis, I see a wide range of applications for good purpose. I really like the examples of detecting terrorism, heresy or diseases that were mentioned.
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteI am thinking of social media posts I terms of sources of such occurrence. Nonetheless, there are many other areas for which these techniques can be applied.
2. Do you think that such tools are safe for our society and freedom of opinion?
No, not really. If somebody can so easily detect what drives us, who we are or what we want then he/she can easily manipulate our opinions for example in any adverts. People may automatically adjust such set up environment and change their opinions. It may be dangerous the aggregate our social media accounts.
3. Could it be used for good purposes?
Probably for education or any other scientific (psychological) research. It helps us understand what drives peoples decisions.
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteThe authors used those data as because it allowed for relatively easy annotation for ideological bias. I think other source might be texts from higlhly biased web sites, choosing only highest-rated ones - readers tend to value opinions in line with their own.
2. Do you think that such tools are safe for our society and freedom of opinion?
I wish a world in which such tools are safe. Where statements are evaluated based on its truth value rather than political affiliation.
The method described applies to english language (no inflecition) in US politics (bi-polar). I think Europe is safe for a while yet from automatic bias detection, as our languages allow to express more nuances in single words, and we are not yet (I hope) stuck with dual-party systems for good. This adds some computational complexity when applying described methods.
3. Could it be used for good purposes?
I think this method has a potential to allow to detect some humorous tones in texts, that is usage of sarcasm and irony.
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteI think the short comments about events connected with politics in specific social media could be good data for similar situation in Poland. There is a lot of people, politicians, journalists and representatives of almost every other profession, who present their opinions on every important events which happens in the country, and surely in such opinions there will be parts involving political views, wich can be detected by this method.
Moreover, in recent days situation in Poland is really focused on the opposite political parties and their influence on mass media. Because of this, a lot of news published by the newspapers or TV programms in fact contains some doses of the political views, and it would be a very efficient field for the research mentioned in the article.
2. Do you think that such tools are safe for our society and freedom of opinion?
As for almost everything – it depends on how we use such tools. First of all – judgment of somebodys opinions is strictly connected with presenting them. So I don’t think that improvement of some tools, which can be helpful with finding political views in some statements, would at some point stop the freedom of speech. When I express my beliefs or comment on something, I take into account that somebody may discover my opinions, including political views. But it is also possible, that for some people knowing that their apparently neutral statements will be examinated to find the true beliefs may stop them from expressing any opinions. Secondly, in my opinion it may have some negative influence on the society, because in Poland people often don’t read all of the presented statement, but only the title. If we tag presentation ( for example liberal / conservative )
people can reject or ignore the whole opinion, which in fact may contains important or interesting views or ideas.
3. Could it be used for good purposes?
For sure this method can be used for good purposes. Most importantly, using such method can be useful for examinate what is the hidden purpose of some statements given by the influential people, for example before the elections. Sometimes it could even expose true views of some persons, when their words and actions are inconsistent.
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteYou mean it in the way of learning datasets? I think that comments (especially: most voted comments) under articles in heavily conservative and liberal websites could be a good input for neural net, that would be already labelled.
2. Do you think that such tools are safe for our society and freedom of opinion?
I think that it mostly depends on the intentions of people who use them and the intelligence level of the receiver (who is reading, for example, content dedicated to the views that he possesses).
3. Could it be used for good purposes?
I think yes, and the good example could be the algorithm which detects if one suffers from depression and potentially prevents suicide (more: https://www.scientificamerican.com/article/can-facebooks-machine-learning-algorithms-accurately-predict-suicide/).
Cheers
1. In the article authors use US congressional floor debate transcripts and ideological books. Could you think of any other kind of data that this method may work for?
ReplyDeleteI guess social media posts or articles would be a pretty good data source.
2. Do you think that such tools are safe for our society and freedom of opinion?
Such tools can never be perceived as safe. If we consider them safe, we would have to consider all people good and I am positively sure this is not the case. Let's be realistic, there will always be one person with bad intentions who would try to take advantage of such a powerful tool.
3. Could it be used for good purposes?
Of course it could, but as I already underlined there will always be at least one person with bad intentions and bad attitude who will think of many other uses of such tools than for "good cause" or "good purpose".
Great blog. All posts have something to learn. Your work is very good and i appreciate you and hopping for some more informative posts. South Florida Divorce lawyer
ReplyDelete