Site Loader

Abstract— Human-centered computing is about developing software beneficial G1 and convenient for users. Media Recommendation systems have been revolutionized since the 1990sG2 , it provides personalized recommendations and predictions over a large group of complex offerings. In this paper, we discussG3 , if the system is really helping to make people improve their choices or affecting user’s choices. G4 G5 The user’s experiences are considered and evaluated for the Media Recommendation System.G6 G7 

Keywords— recommendation system, user study, evaluation, user experience, choice satisfaction.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

I.                    Introduction

 

Human-centered computing G8 plays an important role in supporting and valuing human’s needs and desires. Computing is growing in all ways ubiquitous, grid, mobile and even the social revolution. But these changes or innovations are not always useful, simple, friendly or natural. Thus, those that have to make our daily life simple are making it more complex and complicated. So, it very important to understand human behavior, while developing the technology. Human-centered computing is defined as combining computer sciences with human sciences for designing of a computer system.
G9 G10 

Information keeps growing at a high rate, making the humans difficult to locate it. Recommendation systems bridges this gap . Recommendation system provides recommendations to users based on their choices. The recommendations are usually given based on G11 users rating history and the present items that have highest predicted ratings . For Example, G12 Youtube.com gives suggestions to users based on the ratings they have given to certain videos. An interaction between the G13 user and of the recommendation system is as follows: First, the user’s choices are considered. The system compares the user’s choices with all the available data in the catalog and categories those that would be appreciated by the user. Those with the highest predicted value are given as recommendations to the user.
G14 G15 

A lot of research algorithms and work has been done in the field of recommendation systems. It has been found that the better the algorithm, the better the recommendations, which in turn leads to the better user experience.G16                                                        

The current paper mainly focuses on discussing user’s experience in media recommendation systems. The paper is organized as follows Section II discusses the Related works in this field. Section III discusses the user surveys conducted, its outcomes. Section IV the conclusions and Future work. Section V References.
G17 G18 G19 

 

II.                 Related works

 

Xiao and Benbasat have proposed a marketing-oriented research G20 framework on recommendation systems. Their system tells us how recommendation systems bring a change in user’s decision making.

Hayes et. Al They have proposed a framework for in operational systems for testing user satisfaction. Their research was primarily based on the algorithm. For a particular user, they would test all the algorithms on a particular user. This is quite different from the normal evaluation process, in which one algorithm at one time should be tested.

Zins and Bauernfeind have constructed a model based on the survey conducted by two travelers on users and a system for finding digital cameras to evaluate the user’s experience. This model shows the system satisfaction and factors influencing it like browsing behavior, trust. One drawback is that it doesn’t explain user experience on objective systems.
G21 

Ozok et al. In their approach a questionnaire has been developed and users need to state their opinions in a descriptive way. The drawback is there is no experimental manipulation of a specific system.

Pu and Chen have created a questionnaire for evaluating users experience with the recommender systems. This model also considers users attitudes, behavioral intentions, and beliefs. G22 However, they haven’t considered any personal and situational characteristics. So, selecting concepts from their framework is difficult.

McNee et al. created a Human Recommender Interaction. This was used in the development of the Recommendation system. The analysis is done and used to determine the appropriate system. G23 

In this paper, we will discuss in detail about G24 two different evaluation frameworks for recommendation systems.

III.               Models and Evaluation

 

In this paper, we will G25 discuss two frameworks suggested by different authors in their respective papers. Our first approach is a unifying evaluation framework for the Recommendation systemG26 , called ResQue (Recommender Systems’ Quality of User Experience). This framework aims at measuring user’s satisfaction with the systems, interface and interaction qualities, system usability, quality of recommended items, qualities. In this framework key user experience variables have been identified and special attention has been given to them for conducting user studies. The questions have been structured in 4 layers involving the four dimensions: the perceived system qualities, users believe, subjective attitude and behavioral intentions. This topology clearly explains how users’ perception of the physical features of a system influences their beliefs, attitudes, and finally behaviors. . The first evaluation layer Perceived system qualities focus on functional and informational capabilities of a recommendation system. Four qualities are mainly focused on this layer recommendation Quality, Interaction Adequacy, Interface Adequacy and Information sufficiency and explicability. The second Evaluation layer beliefs focus how effectively and efficiently the system helps users fulfill their tasks. The third layer attitude, it is an overall feeling how users feel about the recommendation system. It is stronger than beliefs. G27 G28 G29 So, if a user shows positive attitude it is taken as user’s satisfaction and trust on the recommendation system. Behavioral Intentions this is related to whether the system influences user’s decisions and motivates them to buy some recommended results. An Experiment was conducted among 239 Experiments. The user’s task was to experience with recommender system by buying an ideal product or just browsing through the system and then to fill up the Resque questionnaire. The diagram below is the results of the structural model analysis.G30 G31 G32 G33 G34 

 

Fig 1: Structural Model Fit

The essential qualities of an effective and satisfying recommender system are determined by the fifteen criteria. There is even both long and short versions of a questionnaire which gives a better scope for designers and researchers. Any recommendation systems including the rating based, utility-based, knowledge-based and also any e-commerce and entertainment websites can be evaluated with this G35 G36 G37 G38 questionnaire.

Another experiment was conducted on paper How G39 Recommender Interfaces Affect user’s opinions 2. Three experiments were conducted on the Movie Lens Recommender system. The experiments are Rerate (Re-rating movie while showing predictions) Unrated (Manipulating predictions for unrated movies) and Scales (Re-rating movies on other scales). The experiment results show that users are consistent while re-rating. In the experiment for users preferred rating scale, it was found that users were more interested in half-star scale most followed by the original movie lens recommendation scale. G40 It is also observed that users are likely to rate same as the original rating when predictions are shown. G41 G42 

  Fig: 2 Percentage of re-ratings below, at, and above

 the original ratings, broken down by whether the original

rating was shown as a prediction.

The results show that people thought their judgment might be wrong and choose the group’s judgment instead. The recommendation systems are usually self-correcting- that is though there are any wrong or artificial ratings, there are other users who give true ratings which makes the system not to recommend it anymore. The results though suggest that as the manipulated predictions have no ratting the self-actualization might be reduced. The results also give ideas to designers about the changes in interfaces. Designers should adopt the scaling system chosen by the users. The users should be given a chance to rate an item when the system shows the item, otherwise, there is a chance that users get manipulated. G43 So, designers to need to keep this in mind while designing. It is observed that manipulations of predictions make the users sensitive. A good algorithm needs to be chosen to incorporate this sensitivity.
G44 G45 G46 

 

IV.               CONCLUSION AND FUTURE WORK

 

Both have frameworks have well explained the user’s intentions. The first framework focuses on user’s attitude and their acceptance of this system. The second framework focuses on how designers need to improve the design so that the G47 users receive the exact recommendations according to their actual interests rather than the manipulated ones. These help in analyzing more users’ perception and improvising according to the system according to their perception. As users desires keep changing it is advisable to perform experiments G48 G49 in a timely manner and incorporate changes accordingly. The future work could be to put more effort in the presentation and interface to improve the usability of the system. More questionnaire like Resque can be done on different groups of people belonging to different parts of the world for more accurate attitudes and behaviors of users while using the recommendation system.G50 G51 G52 

Inserted: –

Inserted: the

Inserted: ,

Inserted: e

Inserted: to

Deleted: about

Deleted:ing

Inserted: –

Inserted: –

Inserted: he t

Inserted: s

Inserted: ,

Inserted:  the

Deleted::

Deleted:ue

Inserted: the

Deleted:about

Deleted:about

Deleted:about

Inserted: –

Inserted: ‘

Inserted: ,

Deleted:n

Inserted: ,

Inserted: ,

Inserted:  the

Inserted: o

Inserted: l

Inserted: o

Deleted:s

Deleted:es

Deleted:i

Deleted:more

Deleted:i

Inserted: –

Inserted: –

Inserted: –

Inserted: the

Inserted: o

Inserted: ,

Deleted:i

Deleted:ly

Inserted: ,

Inserted: s

Deleted:e

Deleted:e

Inserted: f

Inserted:  to

Inserted: o t

Deleted:t

Deleted:,

Deleted:s

Post Author: admin

x

Hi!
I'm Sonya!

Would you like to get a custom essay? How about receiving a customized one?

Check it out