Getting Hooked

As described by Seaver (2019), recommender systems are being designed to captivate or

“hook” users by inclining them to continue favoring more of what they engaged with before.

This makes the user “trapped” in a loop of content they are being recommended on behalf of their consumption and keeps them consuming this content (ibid). This terminology of “hooked”

users was introduced by Eyal (2014) who define it as “successful products reach their ultimate goal of unpromoted user engagement, bringing users back repeatedly, without depending on costly advertising or aggressive messaging” (Eyal, 2014) This process has four stages from trig-ger, to action, the reward and the investment.

Respondent 2 is accessing Ekstra Bladets website many times on a daily basis, as this news media has hooked him in a way that triggers him to take the action of checking articles. By receiving his aim of reading news as presented in the section of News Media Habits he gets rewarded and thereby continues reading articles on the news media site. Finally, he is invested hence he has put time and effort into reading news on Ekstra Bladets digital platform and is likely to return into the loop the next day and repeat this process.

4.3 Ethical considerations

In relation to recommender systems ethics are unavailable. The implementation of AI brings both expectations, but also some concerns amount the consumers of high priority that are pre-sented in the following section.

4.3.1 Personal data collection

During the interviews with the respondents, their perspectives on the collection of personal data were diverse. Hence there was not a conclusive answer to if the respondents found it prob-lematic or not that their data were being collected, but most of them took the collection of per-sonal data for granted. Moreover, the tendency of worrying about data collection or finding it

‘creepy’ was related to the respondents’ gender more than the consumer segment they were a part of. In this part of the analysis, the users’ perception of personal data collection has been analyzed to enable an understanding of the respondents' contrary perspectives.

During the interviews with Respondent 1, Respondent 2 and Respondent 5 they showed indif-ference in connection to the topic of data collection and expressed their acceptance of their data being used: “I do not care. It does not touch me. That’s just the way it is in this world we are living in.” (Respondent number 5, 2021). The three male respondents across the two segments all take it for granted that their data are being collected in the society which they are a part of and believe that is ‘just the way it is’ (Appendix 7). In addition, Respondent 1 mentions his feel-ing in relation to his personal data gettfeel-ing collected as part of the current COVID-19 situation:

“I have already given all my information to private companies who should be allowed to graft me in the nose and stab me in the arm. I have no privacy anymore, it's sold now.” (Respondent number 1, 2021).

The tendency among the male respondents was that they were aware of their data being used in all kinds of scenarios and accepted that it is a part of how society works today. As Respondent 1 verbalizes in the statement, he does not have concerns about a recommender system on Eks-tra Bladet collecting his data as he believes that his privacy on the web is already sold. Ricci et al. (2015) argue that people are willing to give up personal information in return for personal gains. This is the case for implementing these systems on Ekstra Bladet; collecting user data to provide a better news experience for the consumers using the platform (Sloth, 2021).

In the interviews with Respondent 3 and Respondent 4, they expressed their concerns about

Respondent 4 acknowledges that she thinks the idea of collecting data is fine, hence data is collected almost everywhere. Her concern revolves around her data being used to affect other platforms than Ekstra Bladet’s website like Instagram and Facebook:

“I would get tired of if data had been collected and then, when I went elsewhere, I would also be recommended all sorts of articles. I go on a platform because I want to read the news and then I go on Instagram to see stories or pictures. Then I do not bother to see the same things.

Or on Facebook or whatever platform it is.” (Respondent number 4, 2021).

Respondent 4 was not interested in her data being shared among other platforms and third parties and wanted it to be limited to the platform she is on at the current moment. In addition, Respondent 4 makes an active choice of browsing on incognito because she is tired of getting advertisements e.g of the same pair of shoes twenty times (Respondent number 4, 2021). By taking action and consistently browsing in incognito mode the respondent feels more comfort-able on being on digital platforms as she feels more in control over the use of her data. This point of view is supported by Respondent 3 that elaborates on the digitalization and also the increased data collection:

“I do not think it is very nice that we have reached a point where you cannot do anything online anymore without your data being collected and used to influence you, and influencing how you should act online. They can point you in the direction they want and they know so much about you. I think that is really paranoia.” (Respondent number 3, 2021).

Respondent 3 believes the data gathered about her is influencing and can affect one's online behavior. She also describes how the use of her personal data makes her paranoid – which in-sinuates her feeling harassed. This refers to Shklovski et al.’s (2014) term of users perceiving data collection as being “creepy”. Shklovski et al.’s (2014) define this concept as “having or causing a sensation of repulsion, horror, or fear, as of creatures crawling on the skin” (ibid.) Respondent 3 has this fear of the algorithms controlling what she sees and gets exposed to which is also indicated in the section Getting Hooked that makes her feel uncomfortable. “I have accepted it and I live with it. But I think it's really creepy” (Respondent number 3, 2021). The

fact that the respondent has an awareness about her data being collected can increase the feel-ing of her private space befeel-ing infrfeel-inged (Shklovski et al., 2014).

The tendency among the female respondents is they acknowledge their data is being used and accept it, but have different concerns about it. As Respondent 6 verbalizes she finds it “a bit problematic” (Respondent number 6, 2021) that recommender systems on Ekstra Bladet will collect and use her data to give her recommendations. For Respondent 3 giving up her infor-mation is paradoxical, because the respondent is afraid of giving control of her data, but at the same time afraid of missing out on what her friends do on social media; “I would never, dare to delete my Instagram and not follow what all my friends are doing there.” (Respondent number 3, 2021). Awad and Krishnan (2006) argue that useful personalization can motivate consumers to partake in online profiling despite privacy concerns which seem to be the case for the female respondents across the two consumer segments.

It can hence be found that the perception of recommender systems using personal data differ from the respondents. All the respondents accept the data being collected no matter the con-sumer segment they are in. Furthermore, there is a tendency that the male respondent is indif-ferent about their data being collected. In contrast to the female respondents who have some concerns as mentioned about their personal data.

4.3.2 Transparency

The respondents raise in relation to creepy data collection the question of transparency when getting exposed to a recommended article. Lindskow (2021) explains that both transparency and explainability are two key components when they are developing their recommender sys-tems:

“We certainly address explainability and transparency a lot. It's even part of our technical re-search approach - to develop methods that allow us to open up the blackbox to these neural networks and understand why they make their recommendations.” (Lindskow, 2021)

This belief is also apparent in the interviews with some of the respondents as they mention the wish of being informed about which news articles that were recommended specific to them.

“I would definitely be most comfortable if it said ‘recommended to you’. So I knew that this news was actually not something that happened today in Denmark today or had novelty at the moment. It is something that somebody else thinks I would find interesting”. (Respondent num-ber 3, 2021).

From the above quote, it is clear that the respondent is getting more comfortable by being aware of her data being used for recommendations and according to Kruse (2021) there is a general rule that suggests if you can explain why someone got a recommendation, then the user is more satisfied (Appendix). He mentions that explainability is an important factor for under-standing an output, which is also what Respondent 3 requests in relation to getting recom-mended news. Some of the other respondents; Respondent 4 and Respondent 5 take it for granted that their data are being collected and that they are exposed to recommended content when using the web. Respondent 4 explains it his way:

“I think you're well aware these days. Even on Facebook, I know it's not random the order of the content I get to see things and the stories I get shown there.” (Respondent number 4, 2021).

In connection to what Goldschmitt and Seaver (2019) refer to as the concept of ‘Little Brother’

Respondent 5 mentions:

“It might be a bit like Big Brother, if you get recommended something like; this is for you, [name]. But in the end, I think I would not care if it said recommended or not recommended.”

(Respondent number 5, 2021). The respondents mention Big Brother as a comparison to get-ting content explicitly staget-ting the user's name and the phrase ‘recommended for you’. This term is connected to the being recorded through surveillance and according to Goldschmitt and Seaver (2019) you lose privacy when having this ‘Little Brother’ monitor your behavior online e.g when watching a movie or playing a song (ibid.) The respondents, on the other hand, do not mind if the news articles are explicitly stated ‘recommend’ or ‘recommended for you’ . Hence he and the other respondent mentioned above are aware that the content they see online is selected for them already.

The respondents are divided in how they want the news articles to be presented when getting

recommender systems. However, some respondents perceived it unnecessary to explicitly state that some articles are recommended, as they are already aware of being exposed to recom-mended content and take it for granted. According to Stohl et al. (2016), there is a gap between viewability and transparency, which he calls the ‘transparency paradox’. This paradox revolves around the fact that there can be too much information visible for the user, which they do not read because they feel it is transparent. This causes the information to be less transparent.

4.3.3 Influence on Filtering

Respondents across consumer segments mention that they want to have an influence on the filtering. Respondent 1 would like to be able to follow or select journalists on Ekstra Bladet:

“After all, I think there are some journalists on EB who are better than others. There are some journalists that I would like to be able to follow, because they write something like Jonas Sahl, for example. He writes primarily about politics.” (Respondent number 1, 2021). The case for Respondent 2 and Respondent 6 is they want to be able to select their zip number or commune and get local news. Hence this would inform them about the area that they live in but also due to work: “As long as you can switch between zip codes, then it could be so nice especially for us who work municipally.” (Respondent number 2, 2021).

Both respondent 1 and 3 mentions the urge to give feedback on the news articles they get rec-ommended – just like the case with an advertisement on social media platforms:

“I think, for example, the function on Facebook and Instagram is pretty good; when I get all these ads, I can say now I'm tired of looking at them. I think I see this too often. It also provides some data for them in terms of how one is affected.” (Respondent number 1, 2021).

The respondent wants to be involved in the recommendation he gets and be able to give his feedback and influence future recommendations that are not in his interest. The same applies to respondent 4 who wants to be able to go back in her search history and delete an article, hence she believes this would be helpful “so that you eventually end up being able to help shape this algorithm yourself.” (Respondent number 3, 2021).