Present bias leads people to choose smaller immediate rewards over larger rewards in the future. Recommender systems often reinforce present bias because they rely predominantly upon what people have done in the past to recommend what they should do in the future. How can recommender systems overcome this present bias to recommend items in ways that match with users' aspirations? Our workshop position paper presents the motivation and design for a user study to address this question in the domain of movies. We plan to ask Netflix users to rate movies that they have watched in the past for the longterm rewards that these movies provided (e.g., memorable or meaningful experiences). We will then evaluate how well long-term rewards can be predicted using existing data (e.g., movie critic ratings). We hope to receive feedback on this study design from other participants at the HUMANIZE workshop and spark conversations about ways to address present bias in recommender systems.
People often select smaller immediate rewards over larger rewards in the future, a phenomenon that is known as present bias or time discounting. This applies to decisions such as what snack to eat [1, 2], how much to save for retirement [3], or which movies to watch [2]. For example, when people choose a movie to watch this evening they often choose guilty pleasures like The Fast and The Furious, which are enjoyable inthe-moment, but then quickly forgotten. By contrast, when they choose a movie to watch next week, they are more likely to choose iflms that are challenging but meaningful, such as Schindler’s List [2].
Recommender systems (RS), algorithmic systems that predict the preference a user would give to an item, often reinforce present bias. Today, the dominant paradigm of recommender systems is behaviorism: recommendations are selected based on behavior traces (“what users do”) and they largely neglect to capture explicit preferences (“what users say”) [4]. Since “what users do” reflects a present bias, RS that rely upon such actions to train their recommendations will prioritize items that ofer high short-term rewards but low long-term rewards. In this way, recommender systems may reinforce what the current self wants rather than helping people reach their ideal self [5].
This position paper for the HUMANIZE workshop proposes a study design to address these topics in the domain of movies. In Study 1, a survey of Netflix users, we investigate: How should a RS make recommendations by asking ordinary users about the rather academic concept of "long-term rewards"? And can long-term rewards be to transform their behavior [9]. Lukof et predicted based on existing data (e.g., movie al. previously explored how experience samcritic ratings)? In Study 2, a participatory de- pling can be used to measure how meaningsign exercise with a movie RS, we ask: How ful people find their interactions with smartdo users want a RS to balance short-term phone apps immediately after use [10]. Howand long-term rewards? And what controls ever, building such reflection into RSs rewould users like to have over such a RS? mains a major challenge because it is un
We expect that our eventual findings will clear how and when a system should ask a also inform the design of recommender sys- user about the "long-term rewards" of an extems that address present bias in other do- perience. It may be that the common apmains such as news, food, and fitness. proach of asking users to rate items on a "5star" scale reflects a combination of shortterm and long-term rewards, and that a dif2. Related Work ferent prompt is required to capture evaluations of long-term rewards more specifically.
It is also an open question how well such long-term rewards can be inferred from existing data.
The third perspective leverages the wisdom of the crowd, by using the collective elicited preferences of similar users with more experience to make recommendations. Recommender systems today tend to use the "behavior of the crowd" as input into their models, in the form of behavioral data of similar users, but largely neglect elicited preferences [4].
Finally, Ekstrand and Willemsen propose participatory design as a general corrective to the behaviorist bias of recommender systems [4]. Harambam et al. explored using participatory methods to evaluate a recommender system for news, suggesting that giving users control might mitigate filter bubbles in news consumption [11]. Participatory design is a promising way to investigate how users want a RS to balance short-term and long-term rewards and the controls they would like to have.
The social psychologist Daniel Kahneman describes people as having both an experiencing self, who prefers short-term rewards like pleasure, and a remembering self, who prefers long-terms rewards like meaningful experiences [6]. Lyngs et al. describe three diferent approaches to the thorny question of how to measure a user’s “true preferences" [5]. The first approach aligns with the experiencing self, the second with the remembering self, and the third with the wisdom of the crowd.
The first approach follows the experiencing self and asserts that what users do is what they really want, which many in the Silicon Valley push one step further to what we can get users to do is what they really want [7].
Social media that are financed by advertising are "compelled to find ways to keep users engaged for as long as possible" [8]. To achieve this, social media services often give the experiencing self exactly what it wants, knowing that it will override the preferences of the remembering self and lead the user to stay engaged for longer than they had intended.
The second approach prioritizes the remembering self, calling for systems to 3. Proposed Study Design prompt the user to reflect on their ideal self.
In this vein, Slovak et al. propose to design In what follows, we propose a study design to to help users to reflect upon how they wish better understand how to measure the longterm rewards of items in the context of movie recommendations. We hope to receive feedback on this study design from other participants at the HUMANIZE workshop and prompt conversations about ways to address present bias in recommender systems more broadly.
Our first study is a survey of Netflix users that we are currently piloting, which addresses two research questions: • RQ 1a: What wording should be used to ask users to rate the “short-term rewards” and “long-term rewards” of a movie? In other words, what wording captures the right construct and makes sense to users? • For long-term rewards: How rewarding was this movie after you watched it?
scale, from "Not at all" to "Very."
We are also interested in understanding what other constructs are correlated with the both short-term and long-term rewards. To this end, we are also asking about related constructs, such as: • How enjoyable was this movie while
you were watching it? • How interesting was this movie while
you were watching it? • How meaningful was this movie after
you watched it? • How memorable was this movie after you watched it? 3.1.1. Study 1 Methods We will ask Netflix users who watch at least one movie per month to download their past viewing history and share it with us. We will ask them to rate 30 movies: 10 watched in the past year, 10 watched 1-2 years ago, and 10 watched 2-3 years ago. Participants will rate each movie for short-term reward, longterm reward, and other constructs that might be correlated with these rewards (e.g., meaningfulness, memorability).
The current wording of our questions is: We are currently piloting all questions us• RQ 1b: How well can a recommender ing a talk-aloud protocol in which particisystem predict the long-term rewards pants explain their thinking to us as they of a movie for an individual using data complete the survey. We are checking to other than explicit user ratings? make sure that the wording makes sense to participants and to identify constructs that are related to short-term and long-term rewards. The constructs that are most closely related to these rewards in the piloting will be included in the final survey, so that each movie will be rated for a cluster of constructs all related to short-term and long-term rewards. For the final survey, we plan to recruit about 50 Netflix users to generate a total of 1,500 movie ratings.
3.1.2. Study 1 Planned Analysis • For short-term rewards: How rewarding was this movie while you were watching it?
To address RQ 1a, we will report the qualitative results of our talk-aloud piloting of the survey wording. We will also report the correlation between how participants rated our measures of short-term and long-term rewards and related constructs (e.g., meaningfulness, memorability).
To address RQ 1b, first we will test how well existing data correlates with long-term rewards. The existing data we plan to test includes: user ratings (from others), critic ratings, box ofice earnings, genre, and the day of the week the movie was watched. One notable limitation of our study design here is that we will not have access to behavioral data like clicks, views, or time spent about movies.
Second, we will create machine learning models to test how well we can predict the long-term rewards of a movie might provide for an individual, as assessed by metrics like precision and recall. Specifically, we will create: • RQ 2a: What preferences do users have for how a movie RS should weight the short-term versus long-term rewards of the movies it recommends? In what contexts would users prefer what weights? • RQ 2b: How would users like to control how a movie RS weighs the shortterm versus long-term rewards of the movies it recommends?
3.2.1. Study 2 Methods Finally, we will also create generalized and personalized models that predict short-terms rewards too, and compare their performance against the models that predict long-term rewards. Our suspicion is that existing data may be more predictive of short-term rewards than long-term rewards, because longterm rewards may require a form of reflection that most existing data do not capture.
tory design exercise with a movie recommender system, in which we ask: would like to have based on a prototype of a to address cognitive biases in intelligent user news recommender system, with a focus on interfaces [13]. addressing the bias of filter bubbles [11].
Today’s recommender systems often priori- David McDonald, Sean Munson, and Alexis tize “what users do” and neglect “what users Hiniker for feedback on this research agenda say.” As a result, they tend to reinforce the and drafts of the position paper. current self rather than foster the ideal self.
This study design proposes to study this with regards to movies. But the same problem References also applies to other domains such as online groceries, where the current self might want [1] M. K. Lee, S. Kiesler, J. Forlizzi, Mincookies while the ideal self wants blueber- ing behavioral economics to design perries, or digital news, where the current self suasive technology for healthy choices, might want to read stories that agree with in: Proceedings of the SIGCHI Confertheir worldview while the ideal self wants to ence on Human Factors in Computing be challenged by diferent perspectives [12]. Systems, CHI ’11, ACM, New York, NY, The methods we propose in this study design USA, 2011, pp. 325–334. are relevant beyond just movies. [2] D. Read, van Leeuwen B, Predicting
We expect that all workshop participants hunger: The efects of appetite and dewill benefit from a lively discussion of how to lay on choice, Organ. Behav. Hum. Deconceptualize and measure user preferences cis. Process. 76 (1998) 189–205. in ways that go beyond the current behav- [3] H. E. Hershfield, D. G. Goldstein, iorist paradigm of prioritizing what users do W. F. Sharpe, J. Fox, L. Yeykelis, L. L. over explicit preferences. Our proposal to ask Carstensen, J. N. Bailenson, Inusers for their explicit ratings and correlate creasing saving behavior through agethese with other data is just one possible ap- progressed renderings of the future self, proach, and we would like to discuss what J. Mark. Res. 48 (2011) S23–S37. other methods workshop participants would [4] M. D. Ekstrand, M. C. Willemsen, Besuggest and how well these apply to other do- haviorism is not enough: Better recommains such as groceries and news. Address- mendations through listening to users, ing present bias also raises philosophical is- in: Proceedings of the 10th ACM Consues: Is it always irrational to pursue short- ference on Recommender Systems, Recterm rewards over long-term rewards? Are Sys ’16, ACM, New York, NY, USA, 2016, users in a position to judge their own long- pp. 221–224. term rewards? How far should computing [5] U. Lyngs, R. Binns, M. Van Kleek, othsystems go in nudging or shoving users to- ers, So, tell me what users want, what wards long-term rewards? they really, really want!, Extended Ab
Finally, present bias is just one of many stracts of the (2018). cognitive biases. We hope that our submis- [6] D. Kahneman, Thinking, Fast and Slow, sion will also contribute to the growing con- Farrar, Straus and Giroux, 2011. versation on how to use psychological theory [7] N. Seaver, Captivating algorithms: Recommender systems as traps, Journal of Material Culture (2018). URL: https: //doi.org/10.1177/1359183518820366. [8] T. Dingler, B. Tag, E. Karapanos, K. Kise,
A. Dengel, Detection and design for cognitive biases in people and computing systems, http://critical-media.org/ cobi/background.html, 2020. [9] P. Slovák, C. Frauenberger, G. Fitzpatrick, Reflective practicum: A framework of sensitising concepts to design for transformative reflection, in: Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems, dl.acm.org, 2017, pp. 2696–2707. [10] K. Lukof, C. Yu, J. Kientz, A. Hiniker,
What makes smartphone use meaningful or meaningless?, Proc. ACM Interact. Mob. Wearable Ubiquitous Technol.
2 (2018) 22:1–22:26. [11] J. Harambam, D. Bountouridis,
M. Makhortykh, J. Van Hoboken, Designing for the better by taking users into account: a qualitative evaluation of user control mechanisms in (news) recommender systems, in: Proceedings of the 13th ACM Conference on Recommender Systems, dl.acm.org, 2019, pp. 69–77. [12] S. A. Munson, P. Resnick, Presenting diverse political opinions: how and how much, in: Proceedings of the SIGCHI conference on human factors in computing systems, dl.acm.org, 2010, pp.
1457–1466. [13] D. Wang, Q. Yang, A. Abdul, B. Y. Lim,
Designing theory-driven user-centric explainable AI, of the 2019 CHI Conference on . . . (2019). URL: https://dl.acm. org/doi/abs/10.1145/3290605.3300831.