Robot-Fish Interaction Helps to Trigger Social Buffering in Neon Tetras: The Potential Role of Social Robotics in Treating Anxiety

The emerging field of social robotics comprises several multidisciplinary applications. Anxiety and stress therapies can greatly benefit by socio-emotional support provided by robots, although the intervention of social robots as effective treatment needs to be fully understood. Herein, Paracheirodon innesi, a social fish species, was used to interact with a robotic fish to understand intrinsic and extrinsic mechanisms causing anxiety, and how social robots can be effectively used as anxiety treatments. In the first experiment we tested the effects of a conspecific-mimicking robot on the fish tendency to swim in the bottom when transferred in a new tank. Here, P. innesi spent a significantly longer time in the upper section of the test tank when the robotic fish was present, clearly indicating a reduction of their state of anxiety due to social stimuli. The second experiment was based on a modification of the dark/light preference test, since many teleost fish are scototactic, preferring dark environments. However, when the robotic fish was placed in the white half of the test tank, P. innesi individuals swam longer in this section otherwise aversive. Social support provided by the robotic fish in both experiments produced a better recovery from anxiety due to social buffering, a phenomenon regulated by specific neural mechanisms. This study provides new insights on the evolution and mechanisms of social buffering to reduce anxiety, as well as on the use of social robots as an alternative to traditional approaches in treating anxiety symptoms.


Introduction
Social robotics is an emerging multidisciplinary field whose aim is to develop life-like robotic systems socially accepted by humans [1][2][3]. Social robots applications comprise several domains including education, therapy, assistive robotics, human-robot coordinated tasks, search and rescue, domestic aspects [4,5]. Since people tend to anthropomorphize social robots, biomimetic design principles are of crucial importance in this field to ensure robots' interaction and acceptance [6,7].
The wide spectrum of application of social robots also includes their use in reducing anxiety and stress effects to improve the quality of life of patients [8,9]. Anxiety disorders produce strong suffering and dysfunctions in 5-20% of children [10,11] and in 18% of adults [12] and include specific or social phobias, panic, post-traumatic dysfunctions, obsessive-compulsive disorders [13]. The use of social robots to reduce anxiety by providing socio-emotional support, has been found to play an important role in different contexts, and especially during paediatric health-care [14,15].
Robotic solutions aimed at designing human-centered systems are often inspired from nature. Since animals are important model organisms to study factors contributing to anxiety [16][17][18][19][20], they can be used to investigate and refine the intervention of social robots as effective treatment. Furthermore, social robotics techniques can be also used to improve the welfare and health-care of domestic animal species [21]. Indeed, animal health-care, animal welfare, food safety and public health have become of primary importance in policy, as pointed out by the European Food Safety Authority (EFSA) [22,23].
The use of robotic fish mimicking conspecifics of P. innesi can be useful to understand which intrinsic and extrinsic mechanisms cause anxiety, as well as how social robots can be effectively used as pathological anxiety treatments. Animal-robot interactions and ethorobotics are advanced biorobotic and bionic paradigms merging robotics with ethology that enable to establish biohybrid social systems useful for multidisciplinary purposes [29][30][31][32][33][34][35][36][37].
Herein, we developed a robotic platform actuating a robotic fish inspired by P. innesi colour and morphology to investigate the biological mechanisms of social behaviours to effectively use robotic social stimuli in reducing anxiety and posttraumatic stress responses.
Anxiety behaviours in P. innesi was assessed by conducting experiments based on the novel tank diving test, a well-established method to assay for anxiety as a consequence of stressful manipulations in fish [38]. When in a new environment, fish tend to spend a significantly longer time in the bottom of the tank. We measured the duration in the bottom of the tank in both P. innesi exposed to the robotic fish, and non-exposed to the robotic fish to quantify beneficial effects of social stimuli.
To further understand the impact of social biomimetic robots in ameliorating anxiety symptoms, we carried out additional experiments based on a modification of the dark/light preference test, an etho-experimental anxiety model measuring locomotor activity of fish in both dark and light environments as an anxiety index [24]. Several genera of teleosts, including the genus Paracheirodon, were found scototactic (e.g. the preferential pattern of exploration for dark environments) [24,39]. Here, we presented a conflict situation by locating the robotic fish in the light environment, naturally aversive for these fish, and measuring the time spent by neon tetras in both environments.
The animal-robot interaction approach we propose would represent an elective strategy to integrate the usefulness of animals as models in anxiety-related studies, with investigation on the role of social robots as possible therapy.

Ethics Statement
The present study adheres to the Guidelines for the Use of Animals in Research [40], and to the 7010-2020-IEEE Rec-ommended Practice for Assessing the Impact of Autonomous and Intelligent Systems on Human Well-Being [41], as well as to the legal requirements of Italian legislation (D. M. 116,192), and EU regulation [42]. All experiments consisted in behavioural tests, and no specific permits are needed in the country where the experiments were conducted.

Animals Rearing and General Observations
Adult fish of the species Paracheirodon innesi were purchased from a pet store in Pontedera (Pisa, Italy). Fish were maintained under laboratory conditions (at 25 ± 1°C, and 16:8 h light:dark photoperiod) in 100 L aquaria that were filled with activated charcoal-filtered water, at The BioRobotics Institute of Sant'Anna School of Advanced Studies (Pisa, Italy). Cultures were constantly aerated by an air diffuser, and the 30% of the water was replaced every third day. Fish diet consisted in a commercial food (Tetram-in® flake food) that was provided twice a day ad libitum. The aforementioned controlled conditions were the same also during experiments. The laboratory was illuminated by overhead fluorescent daylight tubes (Philips 30 W/33). Diffused laboratory lighting were used to reduce reflection and phototaxis. In each experiment, the behaviour of P. innesi was directly recorded by an observer. Test tanks and the robotic fish were accurately washed at the end of each replicate [32], to avoid effects due to olfactory cues from previous tests. All P. innesi individuals were tested only once.

Robotic Fish and Experimental Apparatus
The robotic fish was designed to resemble P. innesi as much as possible (Fig. 1a), and included a dorsal fin, a second dorsal fin, an anal fin, a caudal fin, two pelvic fins, and two ocular areas. Four parts for each sagittal section of the fish were designed in SolidWorks (Dassault Systemes, Velizy-Villacoublay, France), and fast prototyped in acrylonitrile butadiene styrene (ABS). The final eight parts were assembled by placing a chiffon fabric rectangle (18 × 3 mm) as fish's sagittal plane, between complementary parts (Fig. 1b). The robotic fish was 4 mm wide, 11 mm tall, and 27 mm long. To mimic the color pattern of P. innesi the robotic fish was painted with non-toxic pigments (Fig. 1c). A standard CIELab colour space coordinates determined using a colorimeter (Nix Pro 2 Color Sensor) was used to record colour measurements of the P. innesi body, and of the robotic fish body. A thin layer of transparent silicone rubber (Dragon Skin) covered the robotic fish (Fig. 1d). This fabrication process enabled to have a finalized robotic fish with an increased biomimetic appearance, also allowing passive body undulations, due to its soft and compliant body (Fig. 1c).
The robotic fish was hinged to a magnetic base by a rod. The magnetic base was magnetic coupled with a trajectory Fig. 1 Development phases of the robotic fish. The neon tetra Paracheirodon innesi (a) was used as model to design the conspecificmimicking robotic fish. The robotic fish consisted of 4 parts for each sagittal section that were assembled by placing a chiffon fabric rectangle between complementary parts (b). The robotic fish was then painted with non-toxic pigments to mimic the color pattern of P. innesi (c), and covered by a thin layer of transparent silicone rubber to increase its biomimetic soft appearance (d). A snapshot from the animal-robot social interaction showing a P. innesi individual swimming close to the finalized robotic fish (e) generator below the test tank that moved directly the robotic fish. The trajectory generator had an operating area of around 400 × 200 mm (path following accuracy 0.01 mm), and included two stepper motors, actuating two sliding axis (i.e., x and y axes), and a microcontroller (Arduino Nano) [35]. Plotted trajectories were sent to the microcontroller after conversion in G-Code (i.e., RS-274). An external processor connected to the microcontroller, was used to manage the plotting and code conversion phases.

Experiment 1: Effect of a Social Robot on the Explorative Behaviour of a Novel Tank
Fish were individually placed in a test tank (400 × 300 × 150 mm; length x width x depth) whose water column was divided in two virtual sections: upper section, bottom section (Fig. 2a).
Since the novel tank diving test is based on evoking anxiety behaviour in fish when in a new environment [38], no acclimatation periods was applied from the transfer of fish in the test tank and the start of the test. Fish were exposed to 2 treatments: (i) robotic social stimuli; (ii) no social stimuli. In the treatment i) the robotic fish was located 30 mm below the surface and moved on an ellip-tical trajectory (semi-major axis 195 mm; semi-minor axis 80 mm), with a velocity of 5 mm/s. For each treatment, the time spent by P. innesi in each section of the test tank was measured, considering as index of anxiety the choice of position in the bottom vs. upper sections [38]. The time needed to start swimming (latency) after the release in test tank was also recorded. The test lasted 15 min. Twenty fish were analysed. Each fish was tested only once.

Experiment 2: Effect of a Social Robot on a Dark/Light Preference Context
Herein, a test tank (400 × 300 × 150 mm; length x width x depth) with half black/half white walls and bottom coloured was used to analyses the effect of the robotic fish on the dark/light preference of fish [19] (Fig. 2b)  Three treatments were proposed: (i) robotic fish in the white half; (ii) robotic fish in the black half; (iii) no robotic fish.
When the robotic fish was presented in one of the two halves, it moved on a circular trajectory (∅ 100 mm), with a velocity of 5 mm/s. For each treatment, we measured the time spent by P. innesi in the black and in the white halves of the test tank, as well as the time needed for the first choice of half. We also focally measured the schooling behaviour of P. innesi individuals towards the robotic fish, defined as the tendency of fish to swim at a distance of at least 5 body lengths from other conspecifics and moving collectively [43]. However, Seghers and Magurran [43] also reported that generally fish are observed to school closer than the afore-mentioned distance, and this was also confirmed by our observations.
After each trial the test tank was rotated to avoid orientation effects. Twenty fish were analysed, and only a single test was carried out for each subject.

Statistical Analysis
Data on the impact of the time spent in different sections of the test tank, the duration of the latency, as well as differences among P. innesi fish exposed to robotic social stimuli and no social stimuli were analysed by using non-parametric statistics (Wilcoxon test, P 0.05), as they were not normally distributed (Shapiro-Wilk test, goodness of fit P < 0.05). Differences in the time spent by P. innesi in the black and in the white halves of the test tank, the time needed for the first choice, as well as the schooling behaviour duration, postexposure to the robotic fish in the white half, the robotic fish in the black half, and no robotic fish, were also not normally distributed. Therefore, we relied on Kruskal-Wallis test followed by Steel-Dwass test (P 0.05) to analyse them. All data were analysed by R software v3.6.1 (Stats Package).

Experiment 1: Effect of a Social Robot on the Explorative Behaviour of a Novel Tank
Robotic social stimuli significantly affected the time fish spent in different sections of the test tanks, and in particular P. innesi individuals spent more time in the upper section when the robotic fish was present compared to when no stimuli were provided (Z 5.032; P < 0.0001) (Fig. 3a). When no social stimuli were presented P. innesi individuals spent more time in the bottom section (Z − 4.964; P < 0.0001) (Fig. 3b).
The time spent in different sections was importantly influenced by social stimuli.
The time spent in the bottom section lasted significantly more when no social stimuli were presented (Z − 4.966; P < 0.0001) (Fig. 3c). The time spent in the upper section lasted significantly more when the robotic fish was exposed, (Z 5.033; P < 0.0001) (Fig. 3d). The latency duration decreased significantly when the robotic fish was presented (Z − 4.126; P < 0.0001) (Fig. 3e).

Experiment 2: effect of a social robot on a dark/light preference context
The presence and location of the robotic fish in the test tank significantly affected the time spent by fish in the black half (χ 2 40.928; d.f. 2; P < 0.0001), and in the white half (χ 2 41.089; d.f. 2; P < 0.0001). The time spent in the black half of the test tank was longer when the robotic fish was in the black half compared to when the robotic fish was in in the white half (Z − 5.396; P < 0.0001). The time spent in the black half of the test tank was longer when the robotic fish was not present compared to when the robotic fish was in the white half (Z − 5.399; P < 0.0001) (Fig. 4a).
Fish spent more time in the white half of the test tank when the robotic fish was in the white half compared to when the robotic fish was in in the black half (Z − 5.396; P < 0.0001). Fish spent more time in the white half of the test tank when the robotic fish was in the white half compared to when the robotic fish was not present (Z − 5.399; P < 0.0001) (Fig. 4b).
The time needed by fish for the first choice of half of the test tank was significantly affected by the presence and location of the robotic fish in the test tank (χ 2 36.864; d.f. 2; P < 0.0001). The time needed for the first choice was longer when the robotic fish was not present compared to when the robotic fish was in the black half (Z − 5.184; P < 0.0001), and when the robotic fish was in the white half (Z − 5.021; P < 0.0001) (Fig. 4c). The duration of schooling behaviours was not significantly affected by the location of the robotic fish in the test tank (χ 2 2.681; d.f. 1; P 0.101) (Fig. 4d). When the robotic fish was in the white half of the test tank the time spent by fish in the white half was significantly longer than the time spent in the black half (Z 5.261; P < 0.0001) (Fig. 5a).
When the robotic fish was in the black half of the test tank the time spent by fish in the black half was significantly longer than the time spent in the white half (Z − 5.396; P < 0.0001) (Fig. 5b).
When the robotic fish was not present the test tank the time spent by fish in the black half was significantly longer than the time spent in the white half (Z − 5.396; P < 0.0001) (Fig. 5c).

Discussion
In this study on animal-robot interaction, we show how social robots can contribute to facilitate current research on anxiety, and provide the evidence that can be potentially exploited to ameliorate anxiety-related disorders. Anxiety is a major psychiatric issue whose aetiology is associated with the interaction of several psychosocial factors which lead to neurobiological and neuropsychological dysfunctions [44]. Social robots, used in several healthcare contexts, can represent a strategic method to treat anxiety [45].
In the first experiment we tested beneficial effects resulting from the presence of a conspecific-mimicking robotic fish when a fish individual was transferred in a new tank. It is well established that a new environment causes a stress in fish whose the indicating-anxiety response is represented by the tendency to locate in the bottom of the tank [38,46,47]. However, when the robotic fish was present in the test tank, P. innesi individuals spent a significantly longer time in the upper section of the test tank, clearly indicating a reduction of their state of anxiety due to social stimuli. Also latency lasted considerably less when the robotic fish was exposed, likely do to the effect of social stimuli from the robotic fish that elicited quicker normal foraging activities [48]. These evidences can be associated with a phenomenon widespread in humans and other animals named social buffering [49]. Such a phenomenon, regulated by specific neural mechanisms [50][51][52], consists in a better recovery from an aversive event due to social support. Results from the second experiment further confirmed this hypothesis. P. innesi, together with other teleost species, is a scototactic fish, preferring dark environments [24,39]. Nevertheless, when the robotic fish was placed in the white half of the test tank, real neon tetras preferentially swam in this section otherwise aversive in different treatments. Furthermore, the time needed for the first choice of half was significantly shorter in presence of the robotic fish in one of the two halves of the test tank than in absence of social stimuli. Also, the schooling behaviour duration was not importantly affected by the location of the robotic fish. In general, fish may be involved in novel object investigation displays when exposed to a new stimulus, and several studies addressed this issue to understand how to manipulate important parameters for social affiliation of fish with robotic agents, including the speed and the trajectory of the artifact, as well as its shape, colour, and flexibility [53][54][55]. However, it should be noted that in the experimental conditions of this study (e.g. isolating an individual of a social species and keep it in a new environment) the novel object investigation behaviour would be strongly inhibited, and neophobia, reduced exploration, or hesitancy would predominate [56,57]. Conversely, in our case the robotic fish had a very clear effect in reducing the anxiety behaviour of fish in these adverse conditions, supporting the idea that the artifact triggered social attraction in animals due to its recognizable biomimetic features, as also reported in previous studies on different taxa interacting with conspecific-mimicking robots [58,59]. Thus, we suggest that the effectiveness of robotic social stimuli in buffering anxiety is of particular relevance, since the robotic fish directly influenced fish individual behaviours by increasing their boldness thanks to a conspecific-like presence.
Our findings on neon tetras are valuable to understand the role of social robots in modulating neural mechanisms of social buffering and anxiety control also in humans, since teleosts present brain regions that are homologous to those of mammals [60][61][62][63]. Few studies in mammals have shown that a decrease of the anxiety status occurring in the presence of a conspecific is associated to a lower activation of  In the box plots are included the median (red line) and their lower and upper quartiles and outliers, and green lines and blue T-bars showing mean and standard error values, respectively. On the right of each box plot, histograms showing data distribution are reported the hypothalamus paraventricular nucleus, as well as the lateral and the central amygdala [52,64,65]. Concerning teleost fish, Faustino et al. [49] accurately dissected the neural mechanisms of anxiety reduction resulting from social stimuli in zebrafish, showing that social buffering produced a specific co-activation pattern, involving the medial zone of the dorsal telencephalic area, the supracommissural nucleus of the ventral telencephalic area, and the preoptic area, that are supposed to be homologues of mammals' brain nuclei [63].
This evidence suggests a remarkable evolutionary stable mechanism in controlling anxiety that can be pragmatically crucial for social robotics engineering design, and for robotic-based anxiety treatments. In future research, advanced robot architectures will be implemented (i.e. the Sense-Act-Modulated-by-Interactions architecture [66]), to further increase the interactive features of the experimental apparatus.
This pioneer study offers basic information to unveil the evolution and mechanisms of social buffering to regulate anxiety in social species. Furthermore, it highlights the key role of life-like robotic agents in attenuating anxiety symptoms in clinical contexts, as well as in animal healthcare and welfare.

Conclusion
The present pilot study provides the evidence that animalrobot interactions can be used to advance current research on anxiety, as well as to investigate the role of social robots as a therapy to ameliorate anxiety disorders. Although humas are aware that robots are not conspecifics, unlike our fish experiment, social robotics often identifies, extracts and models principles of human-human interaction to mirror them in robots, using human communication channels, emotion expression, etc., to interact with humans. So, this research on animal-robot interaction provide useful insights to reflect this process in humans. Social support provided by the robotic fish mimicking a P. innesi conspecific produced a better recovery of normal behaviours in real P. innesi individuals when in aversive contexts producing anxiety responses. This phenomenon, named social buffering, is regulated by specific neural mechanisms, and social robots can be used to modulate them to control anxiety.
These findings open a new research avenue for investigating the evolution and mechanisms of social buffering to reduce anxiety, as well as for using social robots as an alternative to traditional approaches to treat anxiety symptoms.
Funding Open access funding provided by Scuola Superiore Sant'Anna within the CRUI-CARE Agreement. Financial support was received from the H2020 FETOPEN Project "Robocoenosis-ROBOts in cooperation with a bioCOENOSIS" [899520]. Funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conflict of interest
The authors declare that they have no conflict of interest.
Ethical approval This study was carried out in accordance with the Guidelines for the Use of Animals in Research, and the 7010-2020-IEEE Recommended Practice for Assessing the Impact of Autonomous and Intelligent Systems on Human Well-Being, as well as the legal requirements of Italian and EU legislation. All experiments consisted in behavioural tests, and no specific permits are needed in the country where the experiments were conducted.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecomm ons.org/licenses/by/4.0/.