Shivam Parikh is a PhD student in UAlbany's College of Engineering and Applied Sciences. He is a systems developer analyst for Information Technology Services who is researching a way to detect fake news. Parikh's mentor is Pradeep Atrey, an associate professor of computer science.
Shivam Parikh is a PhD student in UAlbany's College of Engineering and Applied Sciences. He is a systems developer analyst for Information Technology Services who is researching a way to detect fake news. Parikh's mentor is Pradeep Atrey, an associate professor of computer science.
The UAlbany News Podcast is hosted and produced by Sarah O'Carroll, a Communications Specialist at the University at Albany, State University of New York, with production assistance by Patrick Dodson and Scott Freedman.
Have a comment or question about one of our episodes? You can email us at mediarelations@albany.edu, and you can find us on Twitter @UAlbanyNews.
Sarah O'Carroll:
Welcome to the UAlbany News podcast. I'm your host, Sarah O'Carroll.
Sarah O'Carroll:
People have been deliberately misleading and making up news for a long time. The prevalence of social media, however, has raised the stakes and made it even more challenging to distinguish fact from fiction. I have with me Shivam Parikh, a PhD student in UAlbany's College of Engineering and Applied Sciences. He is a systems developer analyst for information technology services who is researching a way to detect fake news.
News recordings:
Well this is the problem that I have, is when they call us fake news or the enemy of the people or whatever else and then they go out and they blatantly lie like this. How is that not fake news?
News recordings:
WhatsApp has found itself at the center of a growing problem here in India. The spread of fake news often with deadly consequences.
News recordings:
Disinformation, fake news, spread much faster and much further than facts. And as a consequence, they're really good for the business of Facebook, Twitter, and Google.
Sarah O'Carroll:
To start off, Shiv, what is fake news?
Shivam Parikh:
So fake news is not just a buzz word. How I look at it, it's anything that contains any kind of information. And it could be news stories or it could be any kind of information that people read through over the internet or any kind of social media platforms they may or may not use. So that is fake news. It's not just text-based, it goes beyond, more than one media. So my research focuses on not just text based content but more than that. Videos, images, all the other integral parts of a news story or an actual information content.
Sarah O'Carroll:
And so I understand even working with Pradeep Atrey, he's an Associate Professor of Computer Science, on how multimedia content specifically as you just mentioned, is manipulated and what should be done to stop it. So I just like to ask, what about fake news interested you from a research perspective and why you chose to take on this project?
Shivam Parikh:
Sure. So the most interesting part about it is small time modality. It's not just one media. Like I said, it's not just text. It's much more complex than that. It's a title, it's a image, a video, audio, and, of course, the text itself. And yeah, so it's multi-modality is the main reason I chose this topic. And that's why it's very interesting to me.
Sarah O'Carroll:
Very cool. And so what I want to know, what were your questions going in? And on the flip side, what were some takeaways, some things that you learned from it?
Shivam Parikh:
So going in, I would, of course, I see it as just a fake news, just being the one news media. But again, as I studied a little bit further, I discovered that it's not just you about the news, it's not just about the traditional media, it's about social media, it's about people sharing things, how people read such a news story or any news content over the internet. You might've read stats about how 80% of the people, Facebook users actually go on Facebook just for news purposes. So it's not just a traditional media, I mean it has diversed so much in the last decade or so. So that's kind of developed that kind of mentality. And so far I have learned that there are several types of fake news types and I have understood how people actually research certain things and how they interpret certain things. And that's what I've learned so far.
Sarah O'Carroll:
So I'd like to ask, what would you want people to take away, if you could boil it down to a couple of things from your research, would it be to broaden your understanding of fake news or what's sort of the heart of your research that you want people to leave with?
Shivam Parikh:
Sure. So I don't think, I mean we can train. So in a sense you're saying what, what to look out for in fake news or things like that. So I don't think we can, humans are capable of kind of detecting it with 100% accuracy, but there are certain things they can totally pay attention to.
Shivam Parikh:
First off is publisher. Who's publishing such story or any content? What kind of platform they are being published on? And also the content, how the content is written? Is this written to completely feed you what you're supposed to be thinking? Or is it driving you to a certain conclusion? So these are kind of indicators of a fake news. Of course, this is not the only indicators, but humans can kind of look at any article or news story. They can easily spot these things. And of course look out for who's the writer of this? Because a publisher publishes it, but also writers are the actually one writing this such thing. So I would say people can watch out for certain things and dig deeper before reacting to such a headline.
Sarah O'Carroll:
So you mentioned that researchers have sorted fake news into seven categories. So I just was wondering if you might be willing to break down what those categories are? And I'm also just interested why those seven and not five or 10 or something other than that?
Shivam Parikh:
Sure. So a majority of the researchers have kind of broken it down into like these seven categories, which are a visual base, user based, post based, network based, knowledge based, style based and stance based. So I'll say a thing or two about each of them.
Shivam Parikh:
A visual based is more focused on how the appearance of the story. User base is more about what kind of audiences it's catered towards. Post based is more off social media style. A post is a image taxpayer, there's an image, there's a headline and then there are users are reacting to it.
Sarah O'Carroll:
Which is where I feel like most people see their fake news on social media.
Shivam Parikh:
Exactly. Absolutely, yep. Then there is a network based which is for a certain network of people. And then there's knowledge based, knowledge base is very interesting. You might see a lot of new stories. There's a cure for this cancer. It's kind of like there are uncertain scientific answers to certain questions and these knowledge base fake news are kind of answering those questions and people take it for granted because there is no right answer. If something has not been discovered fake news writers can kind of take that up and say, Oh yeah, there's this answer." And that is also fake content. So that goes beyond, this is what the beyond part kicks in for. Just not just the traditional media. Style based is more like focused on how a publisher writes a story in a certain way. It's style of writing. Because if you can detect that that you can kind of detect how certain fake news stories are written and who actually wrote them.
Shivam Parikh:
And then stance based kind of builds upon style based but it's more about not how the story was written, but what that story is kind of delivering. Some of their real stories are more like they give you facts and figures and allows readers to kind of drive the conclusion. Some of the fake news stories are more like the feed you the content. They really direct you in that one exact direction where they want you to think. So stance basis, kind of detecting certain things like that. So that's why these seven categories are kind of important. To kind of understand how the news roam around and over the internet.
Sarah O'Carroll:
It's kind of crazy that there are these seven which just branch out into all sorts of disturbing ways that fake news can be circulated. Another question that I have in all of this is what can be done for media outlets that historically have been considered reputable entities with a tradition of journalistic integrity. But we're seeing them now be berated as carriers of fake news. CNN and The New York Times were just two examples of news organizations that come to mind that have just been attacked for being fake news. And so what can readers do to filter through all of these accusations? And I just feel like this is really important for our democracy, that some semblance of trust remains in the press because if not, we're in a really dangerous place.
Shivam Parikh:
So readers are in a tricky situation because who do they listen to now? Traditional media are also being accused of certain things. But the key part is it's a differentiation between opinion and statement. So a lot of these fake news are kind of opinionated. And readers derive to conclusion to whatever they want to derive too. So I think that's main thing for readers to kind of understand. If let's say some media outlet is saying something, is that an opinion or statement? Because journalists are, sometimes are posing their opinions and readers kind of take that as a statement. So I think they're very, so media outlets are very smart about it when they're speculating certain things, but also at readers will have to be smarter in terms of understanding if it's an opinion or a statement. Don't take opinions as a real content, but look at the statements a little closely.
Sarah O'Carroll:
So it seems like readers should be willing to ask themselves, "Okay, is this piece trying to persuade me? Is this scientific article that seems a bit fishy in some areas? Is it trying to make me do X, Y, Z in my life?" And ask, where's the bias in that?
Shivam Parikh:
Absolutely. Yeah. That should be the mindset for a reader. Yeah.
Sarah O'Carroll:
And at the end of the study, you discuss the four key challenges in fake news detection that will guide future research. And so can you tell me about what those limitations are in existing technologies? Where do we stand now and what's ahead for us?
Shivam Parikh:
Sure. So I'll start with four easy steps that any researchers kind of starts off with and then we'll dive into those challenges. So any researcher, when they're pursuing a problem, they will start off with understanding the problem, searching existing solutions, and then three, implementing their own solution, and four comparing their solutions against other solutions out there. And that is how you kind of push the state of the art to the next level. Now what happens in the fake news research, and what I have studied so far at the four challenges that I've found is, are as following a multimodal the dataset.
Sarah O'Carroll:
And that was going to how it can mean all of these different categories, different ways of fake news can be Russian.
Shivam Parikh:
Right. So yeah, of course, that uses that. But also it's about having a collection of fake news that is not just tech space, but like I said, researchers will have to look at it from a multimedia standpoint. So having the collection of data, a huge dataset of news collection of that contains headline, image, video, any kind of multimedia texts, sources of that story. So not just having the content of the story but the story and the whole. I think having a data set of that will be very helpful for researchers to kind of use that for running their own solution on. So that is one of the challenge that I have come across.
Shivam Parikh:
Another one is a multimodal verification method because there are a lot of work has been done in terms of verifying the text content of the story, but there hasn't been work done if the image has been temporary or that image is actually relevant to that topic or not. So I feel like there needs to be work done in that area.
Sarah O'Carroll:
In other words, taking the multimedia package as a whole, then seeing whether or not that's fake news, right?
Shivam Parikh:
Take a every single part of the story apart, evaluate it on its own merit and then put it back together to kind of find out if they all belong together or not. So yeah.
Shivam Parikh:
And then source verification I like we mentioned earlier who's publishing it, what's their reputation like. If cnn.com is publishing something or New York times is publishing something, let's give them a little bit more credibility given their standing between citizens.
Shivam Parikh:
And then author credibility check. This is an interesting way of looking at certain things because now we know on Facebook, Instagram or Twitter, there's always a blue tick next to somebody who's verified. And I think they needs to have some kind of similar system for story writers to have. Not every journalist went through proper training and schooling. So in this era, anybody can open up a blog and start writing things about it. So I think there needs to be an author credibility check. This goes beyond just verified account. But I think these are the four challenging areas that anybody in fake news research area finds.
Sarah O'Carroll:
I thought that the last one was particularly interesting because you can even be the, we have the problem of not even real humans on Twitter or other social media platforms who are pushing out information. It's automated or whatever. So that does seem like a big need.
Shivam Parikh:
Right, right. And I even think beyond that, that there should be collection of stories written by let's say Sarah, the stories that she has written. I should be able to kind of open that database up and be like, "Okay, let's check how well she's writing things, right?" Because you don't want to just read one story of somebody's and kind of come to a conclusion. So it goes a deeper, depending on the scenario.
Sarah O'Carroll:
Well, we've talked about what research has been done. You just mentioned the next steps for what needs to be done in stopping fake news. But another question for me has to do with us as the American people and whether we can be trusted to distinguish truth from falsehoods or are we willing to just believe a headline that appears in our feed that is paired with a really compelling photo. It's the president with another distinguished world leader or something. And we just want to believe it and we move on in our lives because most likely what's going on in our feeds might already cater to our worldviews because of algorithms and the people in the sites that we follow. So my question is at what point does a journalist's due diligence end and it's up to the reader to exercise democracy and have a more critical eye. I just want to have faith that we are willing to be critical and we're not just going to be mindless consumers.
Shivam Parikh:
Sure. So I mean, so like I had mentioned earlier, it is very hard to train a human brain to kind of think in a certain way because there's a psychological part that kicks in and then people believe what they kind of see. And then they-
Sarah O'Carroll:
Or want to believe.
Shivam Parikh:
Absolutely. And then their friends on social media are commenting. Of course, they will take that for granted because their friends are, let's put them in a verified category. They are, "Oh my friend did say something." So I think it gives, builds that certain amount of trust. So it is, yeah, it is hard for a human to kind of detect a 100%. And that is why researchers will have to kind of create tools that allow readers to kind of think on their own, not just tell them as is fake or true, but kind of give them some kind of metric to decide upon. So yeah, like I said, I don't think it's possible for a human brain to kind of train themselves, that's why there are tools to be created that allows them to come to a right conclusion. There are some people I come across who are not on social media at all, not even LinkedIn. So they try to shut themselves out. But I don't think that's the right way of doing certain things.
Shivam Parikh:
I mean, so it does work. The reason I said it's very hard for a human to detect certain things because you're not going to read a story and then going to research about that publisher. Or you don't have time to go research about what that writer actually wrote other than that. Machine learning can actually do that for users. So I think the tools that would start coming out in future kind of would do certain things for you to kind of save your time. So I think not that human is not capable, but it's that the amount of time they would have to invest to kind of investigate one story. I don't think humans have that time. So that's why they will have to rely on some kind of tools.
Shivam Parikh:
And then the three pillars that I just mentioned earlier, who's the publisher? What kind of platform it's on? Is it verified or what kind of content it is? So they can use these three things. But again, they should, they can investigate on their own. But again, it will take lot of time. That's why I think automated tool would be a great help.
Sarah O'Carroll:
Thank you. And going back to why you originally took on this study to begin with, I want to ask if this survey has changed your understanding of the topic or perhaps even your goals for and with your PhD.
Shivam Parikh:
Yeah, it certainly has. I mean going in starting this topic, fake news to me was just like any news media. Because again, I started the PhD in the time of election time, election 2016, so again, the fake news was a big buzz around that time. So going in I was more about just, Oh, political news and traditional news. But as I learned a little bit more about it, it's a lot more than that. Like going through those seven categories, it goes beyond just a traditional news, political news, its knowledge and how people write stories, how they are catered to a certain audience. So yeah, it has changed my view a significantly and now how I'm targeting to do certain things are more about creating tools that help users. And so I have shifted my game plan in a way, after learning, doing this extensive study on now fake news detection.
Sarah O'Carroll:
When you say change your game plan, as in broaden the scope of what you want to look at or-
Shivam Parikh:
Absolutely. Yeah. Yeah. I was thinking about creating tools for just the news stories and detecting certain types of new stories. But now it has broadened it in a sense, I want to do little bit more than just new stories. I want to do something for social media or for any readers or any content that you may see on the screen. So, yeah, in that sense, it has broadened the scope.
Sarah O'Carroll:
All right. Very cool. Shiv, thank you so much.
Shivam Parikh:
Thank you. Thanks for having me.
Sarah O'Carroll:
Thank you for listening to the UAlbany news podcast. I'm your host, Sarah O'Carroll. And that was Shivam Parikh, a PhD student in UAlbany's College of Engineering and Applied Sciences, and a systems developer analyst for information technology services. This was our pilot episode, and we'd love to hear from you. You can email us at mediarelations@albany.edu or you can find us on Twitter at UAlbany News.