Do Not Let the Robot Get too Close: Investigating the Shape and Size of Shared Interaction Space for Two People in a Conversation
Round 1
Reviewer 1 Report
Lots of grammar mistakes.
What is the influent of robot height on the research? That robot is around 500mm tall, but most of the robots that interact with the human are more then 1m tall? Would it be a difference? What would be if the people will stand instead of sitting?
The description of the kinds of the spaces and people location should be shown on figures to clearly understand the differences. For instance what is the difference between I and L shape formations. The difference in degree is described later on in line 218 (45deg)?
In my opinion figure 4 presents rather V shape chairs location than L shape. It should be more specified by the angle definition or what would be if chairs are positioned parallel? I do not see the explanation of position selection? . However authors describe that orientation is 45deg but in my opinion it is not L (L is 90deg).
What is the head of two people talking position? Does the chair location influence on that. When people talk, they always face to each other, so in I and L shape their sights are pointed in the same directions.
Sections 2.2, 2.2.1 and following consist abbreviations that are not explained in the text of the article.
What does it mean for instance: " F(4,86) = 4.21, p = .004, w2 = 0.22"
I needs description and explanation for the reader.
Figures 2,3,5,6,8,9 do not have units.
Very simple and not scientific conclusion and impact of the research at the domain.
I suggest to add additional experiments and to review the descriptions and figures.
Author Response
Dear reviewer,
Thank you for the review of the manuscript 'Don't let the robot get too close: Investigating the shape and size of shared interaction space for two people in a conversation.' We have carefully considered all comments provided by you and the other reviewer, and have made changes in the manuscript accordingly. All textual changes are highlighted in the newly uploaded text, and below an explanation is provided on how we approached each of your comments. Each comment is marked with a C, and our response with an R. We hope that the revised manuscript and the descriptions below take away most of your concerns regarding the paper, and look forward to receive your feedback.
C: Lots of grammar mistakes.
R: We went through the whole manuscript and found quite some minor grammar issues that we changed accordingly. Major changes in sentence or paragraph structures are highlighted in the new version of the manuscript.
C: What is the influent of robot height on the research? That robot is around 500mm tall, but most of the robots that interact with the human are more then 1m tall? Would it be a difference? What would be if the people will stand instead of sitting?
R: This is a good point, and we did not take this into account in the current studies. The notion that the same robot was used throughout the three studies on one hand makes the results easy to compare, but at the same time limits the generalizability of the work. We have added a section on this in the discussion of the manuscript.
C: The description of the kinds of the spaces and people location should be shown on figures to clearly understand the differences. For instance what is the difference between I and L shape formations. The difference in degree is described later on in line 218 (45deg)?
R: We have included figures that schematically provide an overview of the experimental setup in the three studies. On these figures, the angle between the chairs is clearly visible. The 45deg orientation was meant to be relative to the horizontal line, such that chairs are placed on +45deg and -45deg relative to this line (leading to an angle of 90deg between the chairs). We have clarified this in the new version, as this indeed was unclear.
C: In my opinion figure 4 presents rather V shape chairs location than L shape. It should be more specified by the angle definition or what would be if chairs are positioned parallel? I do not see the explanation of position selection? . However authors describe that orientation is 45deg but in my opinion it is not L (L is 90deg).
R: We already hinted to this in the previous comment, and have made sure that all descriptions in the new version more clearly specify the angle relative to the other chair and/or the horizontal line. The added figures are also designed to provide a clear overview of the placement of the chairs relative to each other.
C: What is the head of two people talking position? Does the chair location influence on that. When people talk, they always face to each other, so in I and L shape their sights are pointed in the same directions.
R: You are absolutely right, thank you for pointing this out. While people are talking, it is indeed expected that they face each other, and we did not limit them into turning their heads in certain positions. We expected the people to look at the robot once in a while, and look at the other participant the rest of the time. While the robot is in their peripheral vision, they would still be able to see the robot while deciding to stop it. When the robot was approaching them from the back, they will probably have made their decision based on auditory signals from the robot. We have expanded the discussion to further elaborate on this issue.
C: Sections 2.2, 2.2.1 and following consist abbreviations that are not explained in the text of the article. What does it mean for instance: " F(4,86) = 4.21, p = .004, w2 = 0.22" I needs description and explanation for the reader.
R: The abbreviations represent the statistical findings. In linear mixed models, the statistics are f-tests with degrees of freedom based on the number of conditions and data points. The p-value indicates significance. Omega squared represents the effect size. This is a common way of reporting the statistics, as can be seen at http://my.ilstu.edu/~jhkahn/apastats.html on reporting statistics and https://www.statisticshowto.datasciencecentral.com/omega-squared/ on the calculation of omega squared.
C: Figures 2,3,5,6,8,9 do not have units.
R: The captions of the figures now explain the units in which the variables are presented.
C: Very simple and not scientific conclusion and impact of the research at the domain.
R: We always aim to provide a conclusion of our work that surpasses the specific context of our experiments, but we agree that the conclusion was indeed a bit vague. We have changed it to better fit the scientific contribution.
C: I suggest to add additional experiments and to review the descriptions and figures.
R: The descriptions and captions with the figures are updated, but adding additional experiments was not possible within the 10-day limit for this revision. We agree that collecting more data would provide valuable insights in the generalizability of the work (as the other reviewer also highlighted), but we also believe that the current manuscript sufficiently describes the notion that people take each others' personal space into account when stopping a robot that approaches them.
Reviewer 2 Report
The authors of this manuscript present a set of user studies to experimentally validate the shape and size of the shared interaction space of two people in a conversation. The manuscript includes useful background and related work in robot proxemics in HRI, shared interaction space and human formation patterns.
The main objective of the three presented studies is to evaluate the different approach direction and stopping distance of a NAO robot while it approaches two human participants who are supposed to be engaged in a conversation. They use different human formation patterns. According to the studies and their results, the paper provides insights on how a robot would approach multiple users.
The manuscript is well-written and provides useful information in this area. However, I would suggest the following edits:
I would reorganize the structure of the paper, grouping all three studies. Since the method, procedure and results sections are repeated for each study, I would have one section for the study setup and its variations, one section for the participants, etc. It would be interesting to present the results of all studies together to compare the different parameters. I believe the study [protocol is a little bit vague in terms of the participant preparation. Were the participants instructed to press the button according to comfort? Maybe a more concrete scenario would help the study be more specific, e.g. what is the role/goal of the robot approaching, for example, to initiate a new conversation with one of the participants or to participate in the current one, etc. Moreover, since when only one subject is available the experimenter participates, how do you use the subject's data. I thought that both users need to press the button and both inputs matter. It would be also really interesting to compare these results for different contexts or robot roles/goals. This could be discussed in the manuscript.
Author Response
Dear reviewer,
Thank you for the review of the manuscript 'Don't let the robot get too close: Investigating the shape and size of shared interaction space for two people in a conversation.' We have carefully considered all comments provided by you and the other reviewer, and have made changes in the manuscript accordingly. All textual changes are highlighted in the newly uploaded text, and below an explanation is provided on how we approached each of your comments. Each comment is marked with a C, and our response with an R. We hope that the revised manuscript and the descriptions below take away most of your concerns regarding the paper, and look forward to receive your feedback.
C: I would reorganize the structure of the paper, grouping all three studies. Since the method, procedure and results sections are repeated for each study, I would have one section for the study setup and its variations, one section for the participants, etc. It would be interesting to present the results of all studies together to compare the different parameters.
R: We agree that combining the studies into a larger section could make it easier to compare the findings. With a 10-day limit we were however not able to fully change the structure accordingly and have decided to keep it as it is.
C: I believe the study [protocol is a little bit vague in terms of the participant preparation. Were the participants instructed to press the button according to comfort? Maybe a more concrete scenario would help the study be more specific, e.g. what is the role/goal of the robot approaching, for example, to initiate a new conversation with one of the participants or to participate in the current one, etc.
R: We have changed the description of the method to include more details on the protocol that was used in the experiments and hope this clarifies these issues.
C: Moreover, since when only one subject is available the experimenter participates, how do you use the subject's data. I thought that both users need to press the button and both inputs matter.
R: When this was the case, the experimenter pressed the button shortly after the robot started walking, without the participant noticing. This clarification is added to the manuscript.
C: It would be also really interesting to compare these results for different contexts or robot roles/goals. This could be discussed in the manuscript.
R: The other reviewer also pointed out that results are not generalizable when only one (type of) robot is used in the experiments. We agree, and we have added a section in the discussion on this.
Round 2
Reviewer 1 Report
All the suggestions were taken into consideration and implemented to the article.
As far as the comparison of the results is concerned the use of the same robot enables easier to compare. Authors added a section with discussion on this.
The corrected figures clarify the path of conducting the experiments.
However the additional experiment will add value to the discussion and comparison but this not disqualifies the article from publication.
Author Response
Dear reviewer,
Thank you for you quick response to the changes in the manuscript. We are glad that the changes satisfied your concerns, and although we agree that adding an additional experiment would add value to the work, time limitations did not allow for adding this. We did edit the introduction and discussion to include more relevant literature, as highlighted in the new version of the manuscript.