CoChat

An adaptive decision-making system supported on user preference predictions for human–robot interactive communication

Open access
Maroto-Gómez, Marcos, Castro-González, Álvaro, Castillo, José Carlos, Malfaz, María, Salichs, Miguel Ángel · 2023 · User Modeling and User-Adapted Interaction

The DOI matches its Crossref record; 72 of the 72 references it cites were found in a registry or an index.

Abstract

Adapting to dynamic environments is essential for artificial agents, especially those aiming to communicate with people interactively. In this context, a social robot that adapts its behaviour to different users and proactively suggests their favourite activities may produce a more successful interaction. In this work, we describe how the autonomous decision-making system embedded in our social robot Mini can produce a personalised interactive communication experience by considering the preferences of the user the robot interacts with. We compared the performance of Top Label as Class and Ranking by Pairwise Comparison, two promising algorithms in the area, to find the one that best predicts the user preferences. Although both algorithms provide robust results in preference prediction, we decided to integrate Ranking by Pairwise Comparison since it provides better estimations. The method proposed in this contribution allows the autonomous decision-making system of the robot to work on different modes, balancing activity exploration with the selection of the favourite entertaining activities. The operation of the preference learning system is shown in three real case studies where the decision-making system works differently depending on the user the robot is facing. Then, we conducted a human–robot interaction experiment to investigate whether the robot users perceive the personalised selection of activities more appropriate than selecting the activities at random. The results show how the study participants found the personalised activity selection more appropriate, improving their likeability towards the robot and how intelligent they perceive the system. query Please check the edit made in the article title.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1007/s11257-022-09321-2

Cites

72 of 72 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 An instance combines the input vector with the labelled output in ranking format. On the one hand, the input vector contains the user’s features regarding the personal, interests, and preferences data about different attributes. On the other hand, the labelled output concerns a predefined set of items (labels) preference ranking. In this case, the example ranks the categories multimedia, games, and information according to the user’s preferences
Fig. 1 An instance combines the input vector with the labelled output in ranking format. On the one hand, the input vector contains the user’s features regarding the personal, interests, and preferences data about different attributes. On the other hand, the labelled output concerns a predefined set of items (labels) preference ranking. In this case, the example ranks the categories multimedia, games, and information according to the user’s preferencesCC BY · PubMed Central
Fig. 2 Mini, a social robot designed for entertainment and assisting older adults in cognitive stimulation therapies
Fig. 2 Mini, a social robot designed for entertainment and assisting older adults in cognitive stimulation therapiesCC BY · PubMed Central
Fig. 3 General view of the architecture of the social robot Mini, highlighting its decision-making system module in charged on sequencing the robot’s behaviour
Fig. 3 General view of the architecture of the social robot Mini, highlighting its decision-making system module in charged on sequencing the robot’s behaviourCC BY · PubMed Central
Fig. 4 Organisation in three hierarchical levels (categories, subcategories, and entertainment activities) of datasets generated by surveying users. Low-level items are ranked at its higher level producing a ranking of preferences for each user input. In brackets next to each label, its corresponding abbreviation
Fig. 4 Organisation in three hierarchical levels (categories, subcategories, and entertainment activities) of datasets generated by surveying users. Low-level items are ranked at its higher level producing a ranking of preferences for each user input. In brackets next to each label, its corresponding abbreviationCC BY · PubMed Central
Fig. 5 Set of tablet menus shown by the robot giving the users the possibility to select the option they prefer
Fig. 5 Set of tablet menus shown by the robot giving the users the possibility to select the option they preferCC BY · PubMed Central
Fig. 6 Flow diagram depicting the full process for predicting user preferences and how the autonomous decision-making system use them to provide personalized activities to different users
Fig. 6 Flow diagram depicting the full process for predicting user preferences and how the autonomous decision-making system use them to provide personalized activities to different usersCC BY · PubMed Central
Fig. 7 Robot autonomy level in activity selection depending on the Case of Study. A non-autonomous robot (Case of study 1) leaves decisions in the user’s hands. The selections produce using menus displayed on the touch screen. The semi-autonomous mode combined user selections with autonomous decisions of the robot using the preference learning system. Finally, a fully autonomous robot always decides the entertainment activity to execute using the preference learning system. The decision-making system selects the operating mode depending on the proactivity level of the user
Fig. 7 Robot autonomy level in activity selection depending on the Case of Study. A non-autonomous robot (Case of study 1) leaves decisions in the user’s hands. The selections produce using menus displayed on the touch screen. The semi-autonomous mode combined user selections with autonomous decisions of the robot using the preference learning system. Finally, a fully autonomous robot always decides the entertainment activity to execute using the preference learning system. The decision-making system selects the operating mode depending on the proactivity level of the userCC BY · PubMed Central
Fig. 8 Results carried out to test the operability of the preference learning system in the social robot Mini. First, we compare Top Label as Class and Ranking by Pairwise Comparison to obtain the algorithm with the best performance in our application. Then, we present three case studies about the operation of the system when interacting with new users with different proactivity levels. Finally, we validate the system in real human–robot interactions to assess whether the personalised selection of entertainment activities is preferred above a random activity selection
Fig. 8 Results carried out to test the operability of the preference learning system in the social robot Mini. First, we compare Top Label as Class and Ranking by Pairwise Comparison to obtain the algorithm with the best performance in our application. Then, we present three case studies about the operation of the system when interacting with new users with different proactivity levels. Finally, we validate the system in real human–robot interactions to assess whether the personalised selection of entertainment activities is preferred above a random activity selectionCC BY · PubMed Central
Fig. 9 Representation of ntrees and mdepth hyperparameters optimization on each dataset considering the set of hyperparameter that provided the best results in each trial
Fig. 9 Representation of ntrees and mdepth hyperparameters optimization on each dataset considering the set of hyperparameter that provided the best results in each trialCC BY · PubMed Central
Fig. 10 Predicted decision tree for User 1 showing how each category, subcategory, and activity rank in the hierarchy according to the score (in brackets) estimated by the preference learning system. Highlighted in red, the path that represents the preferred category, subcategory, and entertainment activity for the user. In this case, the category is Multimedia, the subcategory is Sayings, and the activity is playing a random saying
Fig. 10 Predicted decision tree for User 1 showing how each category, subcategory, and activity rank in the hierarchy according to the score (in brackets) estimated by the preference learning system. Highlighted in red, the path that represents the preferred category, subcategory, and entertainment activity for the user. In this case, the category is Multimedia, the subcategory is Sayings, and the activity is playing a random sayingCC BY · PubMed Central
Fig. 11 Predicted decision tree for User 2 showing how each category, subcategory, and entertainment activity rank inside the hierarchy according the score of each item. Highlighted in red, the path indicating the activity breaking news (BN) contained in the Information category that the user likes the most
Fig. 11 Predicted decision tree for User 2 showing how each category, subcategory, and entertainment activity rank inside the hierarchy according the score of each item. Highlighted in red, the path indicating the activity breaking news (BN) contained in the Information category that the user likes the mostCC BY · PubMed Central
Fig. 12 Predicted decision tree for User 3 showing how each category, subcategory, and activity rank inside the hierarchy according to its associated estimated score (in brackets). Highlighted in red, the path indicating which activity (classic games (CG)) has more possibilities to be chosen in case the robot decides the activity autonomously
Fig. 12 Predicted decision tree for User 3 showing how each category, subcategory, and activity rank inside the hierarchy according to its associated estimated score (in brackets). Highlighted in red, the path indicating which activity (classic games (CG)) has more possibilities to be chosen in case the robot decides the activity autonomouslyCC BY · PubMed Central
Fig. 13 Mean and standard deviation value of the Personalisation perceived attribute for Condition 1 (C1): a robot personalising the activity selection and Condition 2 (C2): a robot selecting activities randomly. Very significant statistical differences (p-value from 0.01 to 0.001) are denoted with ***
Fig. 13 Mean and standard deviation value of the Personalisation perceived attribute for Condition 1 (C1): a robot personalising the activity selection and Condition 2 (C2): a robot selecting activities randomly. Very significant statistical differences (p-value from 0.01 to 0.001) are denoted with ***CC BY · PubMed Central
Fig. 14 Mean and standard deviation values in the statistical analysis of the Godspeed results considering each category and condition tested (Condition 1 (C1): a robot personalising the activity selection and Condition 2 (C2): a robot selecting activities randomly). Anthropomorphism (A), Animacy (A), Likeability (L), Intelligence perceived (IP), and Security perceived (SP). Significant statistical differences (p-value from 0.05 to 0.01) are denoted with *, very significant (p-value from 0.01 to 0.001) with **, and extremely significant (p-value<0.001) with ***
Fig. 14 Mean and standard deviation values in the statistical analysis of the Godspeed results considering each category and condition tested (Condition 1 (C1): a robot personalising the activity selection and Condition 2 (C2): a robot selecting activities randomly). Anthropomorphism (A), Animacy (A), Likeability (L), Intelligence perceived (IP), and Security perceived (SP). Significant statistical differences (p-value from 0.05 to 0.01) are denoted with *, very significant (p-value from 0.01 to 0.001) with **, and extremely significant (p-value<0.001) with ***CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 72 references

Research works printed but not linked: 13

  1. Aiolli, F., Sperduti, A.: Learning preferences for multiclass problems. Adv. Neural Info. Process. Syst., pp 17–24 (2005)
  2. Brinker, K., Hüllermeier, E.: Case-based multilabel ranking. In: IJCAI, pp 702–707 (2007)
  3. Burke, R., Felfernig, A., Göker, M.H.: Recommender systems: an overview. Ai Mag. 32(3), 13–18 (2011)
  4. Chen, Y., Zhang, J., Guo, M., Cao, J.: Learning user preference from heterogeneous information for store-type recommendation. IEEE Trans. Serv. Comput. (2017)
  5. Cohen, W.W., Schapire, R.E., Singer, Y.: Learning to order things. Adv. Neural Info. Process. Syst., pp 451–457 (1998)
  6. Dery, L.: Multi-label ranking: Mining multi-label and label ranking data. arXiv preprint arXiv:210100583 (2021)
  7. Efron, B.: Student’s t-test under symmetry conditions. J. Am. Stat. Assoc. 64(328), 1278–1302 (1969)
  8. Goztepe, K.: New directions in military and security studies: artificial intelligence and military decision making process. Int. J. Info. Security Sci. 4(2), 69–80 (2015)
  9. Olaronke, I., Oluwaseun, O., Rhoda, I.: State of the art: a study of human-robot interaction in healthcare. Int. J. Info. Eng. Elect. Bus. 9(3), 43 (2017)
  10. Prelipcean, G., Boscoianu, M., Moisescu, F.: New ideas on the artificial intelligence support in military applications. In: Proceedings of the 9th WSEAS international conference on Artificial intelligence, knowledge engineering and data bases, World Scientific and Engineering Academy and Society (WSEAS), pp 34–39 (2010)
  11. Woodworth, B., Ferrari, F., Zosa, T.E., Riek, L.D.: Preference learning in assistive robotics: Observational repeated inverse reinforcement learning. In: Machine learning for healthcare conference, PMLR, pp 420–439 (2018)
  12. You, R., Zhang, Z., Wang, Z., Dai, S., Mamitsuka, H., Zhu, S.: Attentionxml: label tree-based attention-aware deep model for high-performance extreme multi-label text classification. Adv. Neural Info. Process. Syst. 32, 5820–5830 (2019)
  13. Zheng, Z., Zha, H., Zhang, T., Chapelle, O., Chen, K., Sun, G.: A general boosting method and its application to learning ranking functions for web search. Adv. Neural Info. Process. Syst., pp 1697–1704 (2008)

Cited by

See all 37

Metadata

Authors
Maroto-Gómez, M., Castro-González, Á., Castillo, J. C., Malfaz, M., & Salichs, M. Á.
Affiliations
Universidad Carlos III de Madrid (ES)
Venue
User Modeling and User-Adapted Interaction
Published
2022-04-09
Volume
33
Issue
2
Pages
359-403
ISSN
  • 0924-1868 (ISSN-L)
  • 1573-1391
OpenAlex
W4224292998
Cited by
37
References
72
Citations by year
2026: 6 · 2025: 9 · 2024: 10 · 2023: 9 · 2022: 3
FWCI
7.32
Type
journal article
Language
en
Open access
Licence
Locations
  • User Modeling and User-Adapted Interaction (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • e-Archivo (Carlos III University of Madrid) (submittedVersion, cc-by)
  • e-Archivo (Carlos III University of Madrid) (submittedVersion, cc-by-nc-nd)
Funders
Ministerio de Ciencia, Innovación y Universidades
Topics
  • Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
  • Physical Sciences › Computer Science › Information Systems › Recommender Systems and Techniques
  • Physical Sciences › Computer Science › Artificial Intelligence › Speech and dialogue systems
Keywords
user preference prediction, social robots, human-robot interaction, adaptive decision-making, preference learning, robot likeability, small robots

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: An adaptive decision-making system supported on user preference predictions for human–robot interactive communication. https://app.cochat.ai/papers/10.1007/s11257-022-09321-2
The paper
Maroto-Gómez, M., Castro-González, Á., Castillo, J. C., Malfaz, M., & Salichs, M. Á. (2023). An adaptive decision-making system supported on user preference predictions for human–robot interactive communication. User Modeling and User-Adapted Interaction. https://doi.org/10.1007/s11257-022-09321-2

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.