1 Introduction

A changing age structure accompanied by an increased need for care and a simultaneous shortage of care personnel pose major challenges for today’s society [1,2,3]. For instance, in Germany more than one in five persons is aged above 65 years and more than two thirds from all people aged beyond 90 years need intensive support and care [4]. At the same time, there is a lack of qualified and adequately trained personnel to care for all these (older) people in need of care (e.g., [5]). Despite these developments, the majority of people in older age desires to stay within their own home as long and as independent as possible. Therefore, innovative approaches must be designed and realized to meet the needs and wishes of (older) people in need of care.

One of these innovative approaches lies in the development of ambient assisted living technologies (AAL). The latter have the potential to support older and people with physical and/or cognitive decline enabling them to stay longer within their own home increasing their safety, well-being, and autonomy (e.g., [6,7,8]). Thereby, the functional potential of those approaches is extremely broad and reaches from health monitoring and cognitive orthetics, over emotional support (e.g., for family or professional caregivers), to the detection of emergencies, such as falls [7,8,9]. These approaches are realized applying a broad range of wearable or ambient sensors and usually they consist of a combination of different technologies. The most relevant elements include active or passive infrared sensors, radio frequency identification, pressure or ultrasonic sensors, magnetic switches as well as audio-based microphone and visual technologies [7, 10]. The latter mentioned video-based or visual AAL approaches are particularly promising, as they are pursued, e.g., to detect or even prevent emergencies such as falls using spatiotemporal information or information based on body posture and head position [11,12,13]. Besides the variety of the technical possibilities and the advantages the usage of such systems brings along, it is necessary to keep social aspects in terms of the perceptions of future users in mind, as such technical approaches deeply intervene in the autonomy of their users. Considering the usage of video-based AAL systems, in particular perceived barriers and concerns with regard to potential invasions of privacy and data security have to be considered, understood, and investigated in detail, as they can represent an obstacle to a sustainable acceptance and adoption of the visual AAL technology in the users’ everyday life [14,15,16,17].

Although there has been significantly more research on the acceptance of AAL technology as well as on perceptions of privacy in the context of AAL in recent years, to the best of our knowledge, specific investigations of privacy preferences have hardly been analyzed focusing on using visual AAL technologies and taking their characteristics into account (e.g., feelings of being monitored). Beyond that, a detailed understanding of privacy perception patterns differing between human and technological invasions is missing. Hence, this study aimed at a detailed investigation of human-, technology-, and context-related factors affecting privacy perceptions when visual AAL technologies are used. Based on these results, concrete recommendations for a user-centered and -orientated design and adaption of visual AAL technology can be derived.

The paper is structured as follows. Section 2 represents the theoretical and empirical background of the study, describing previous research on the acceptance of (visual) AAL technology, the role of privacy for technology acceptance, specific influencing factors on the perceptions and acceptance, as well as the aim and underlying research questions of the present study. In Sect. 3, the methodological approach of the study is presented focusing on the design of the applied online survey, data analysis, and a description of the sample. Subsequently, Sect. 4 presents the results of the study based on the previously introduced research questions. Finally, the results are discussed in Sect. 5 enabling to derive concrete recommendations for the design of user-centered and privacy-aware visual AAL technology.

2 Technology acceptance and the relevance of privacy

In the following, the current state on the acceptance of (visual) AAL technology is presented. Thereby, the role of privacy is highlighted and relevant definitions, previous research, and influencing factors are described.

2.1 Research on AAL acceptance and perceptions of privacy

Previous research on the acceptance of AAL technologies frequently revealed positive assessments by future users. Thereby, the usefulness of technical support in older age as well as its potential as measure for the challenges in healthcare due to demographic change are acknowledged [18, 19]. To investigate the acceptance of AAL technologies, the well-known and established models of technology acceptance TAM [20] and UTAUT [21] are frequently applied and adapted [22]. However, these models allow a rather generic investigation of acceptance patterns omitting specific technology- or context-related acceptance parameters. Therefore, more explorative, and specifically tailored empirical approaches are needed.

In this regard and beyond the application of the acceptance models, previous research has already identified that the acceptance of AAL technologies and systems is impacted by the perception of respective benefits and barriers (e.g., [23,24,25]). Considering perceived benefits and motives to use AAL technology, an increased independence and autonomy, feelings of safety, and staying longer within the own home impact the acceptance positively [23, 25]. However, it is even more important to consider perceived barriers and concerns regarding the usage of AAL technologies, since they have the potential to hinder or decrease a sustainable acceptance and adoption: Here, an invasion of personal privacy, feelings of surveillance and control, and isolation due to a reduction of human contact and care have been confirmed to be relevant parameters [23,24,25,26]. Beyond that, an unauthorized access for third parties and a felt loss of control of sensitive data represent essential barriers as well [27]. These barriers become even more important when visual AAL technologies are used: In this specific regard, previous research identified that visual camera-based technologies are less accepted and preferred in comparison with other AAL technologies, such as emergency buttons or smart watches [28]. Thereby, the main reasons for the unwillingness to integrate visual AAL technologies in the own home environments predominantly were attributed to privacy concerns (e.g., [29]). Hence, these previous research results show that privacy-related concerns can play a major role for the acceptance of visual AAL technologies and have to be understood in detail taking specific functions, technologies, and contexts into account.

Privacy and the perception of privacy represent complex and multi-faceted phenomena (e.g., [30,31,32,33]). One established definition describes privacy as the control over access to oneself (e.g., [31, 34, 35]), aiming at an achievement of the optimal balance between withdrawal and disclosure in each context-bounded situation [31, 33]. A model of Burgoon [36] defines four dimensions of privacy: informational privacy, social privacy, physical privacy, and psychological privacy. Taking the development of digital technologies into account, informational privacy describes the control over access to personal information; while, social privacy relates to the control over personal interactions with others. Physical privacy describes control over access to one's person, such as freedom from surveillance and undesired intruders. Finally, psychological privacy means the protection against interference with one’s thoughts, values, and emotions [36]. As the use of AAL technology in the everyday life has the potential to invade these four dimensions, they have to be considered and investigated in order to do justice to future users’ requirements. In fact, an invasion of privacy, i.e., when these control mechanisms in the different dimensions fail to function, can lead to the development of emotional disorders such as mood or anxiety disorders, stress or even depression which underlines the importance of studying privacy issues in this context (e.g., [37,38,39,40]). As suggested above, privacy concerns have already been identified to represent relevant barriers for technology acceptance [41], in particular regarding AAL technology [14, 23, 29]. In general, previous research identified that privacy concerns are influenced by the specific technology itself, e.g., technology type and location of technology application [16], by the sensitivity of the collected data [42], and by the person/institution having access to the collected data [43]. However, these findings relate to health technology or AAL technology rather generically. Focusing on visual-based AAL technology, there is currently hardly any knowledge about such specific privacy perceptions and requirements. Hence, it must be investigated which data are perceived as most sensitive, which concrete privacy concerns exist, which specific visual technology is least privacy-invasive and most preferred, or which feelings are associated with the usage of visual AAL technology.

2.2 Influencing factors on acceptance and privacy perception in using AAL technology

Research regarding perceptions of privacy has increased over the last decades investigating different impacting parameters. Starting with human-related factors, diverse studies identified age, gender, and health status to be impacting factors for the overall perception of privacy and data security (e.g., [44, 45]). Those studies focused on an overall evaluation of the relevance of privacy as well as on relevant privacy requirements. Other qualitative approaches [46] investigated a comparison what privacy meant to the participants in a technical and a non-technical context, and identified different existing mental models of privacy and privacy violations: Thereby, some participants argued that private data can be used in a good and in a bad way, while others completely rejected to use technology due to fears of privacy protection and negative feelings toward services who would use private data.

As there are so far no direct comparisons of human and technological privacy invasions and violations, these differences serve as a starting point to analyze the perceptions of human and technological invasions of privacy in more detail, allowing direct quantitative comparisons regarding affective evaluations and perceived concerns (see RQ1 & RQ2).

Considering technology-related factors within the perception of privacy, previous research has already shown that visual technologies are perceived as more privacy intrusive and are less accepted compared to audio- or sensor-based technologies (e.g., [16, 28]. As these results referred to conventional cameras, it is vague and must be investigated, which visual technology would be preferred in direct comparisons (e.g., conventional vs. depth camera) (RQ4). Taking context-related factors into account, previous research identified the locations in which AAL technology is installed as a privacy- and acceptance relevant factor (e.g., [16]). Here and in addition to the locations of video-based monitoring, detailed insights into the perceptions of visual technologies are missing, in particular in terms of what is visual AAL technology allowed and not allowed to “see,” to monitor, and to record (RQ3). Beyond that, previous studies revealed that medical necessity in terms of care impact the perception and acceptance of assisting technology [28]. Here, it should be investigated to what extent different contexts of monitoring (e.g., daily activities vs. caring context) impact the evaluation of visual AAL technology as well as the selection of preferred visual technologies (RQ3 & RQ 4). Addressing these human-, technology-, and context-related factors, this study contributes to closing the research gap regarding the perception of privacy requirements when visual AAL technologies are used.

2.3 Research aim and questions

The present study aimed at gaining detailed insights into influencing parameters on the perceptions and preferences of privacy—when visual AAL technologies are used. Thereby, it had to be investigated whether and to what extent privacy preferences differ comparing human and technological invasions of privacy. In a second step, the impact of different (privacy) contexts had to be analyzed regarding the acceptance of visual recordings of activities of daily living focusing on which daily activities are (not) allowed to be seen and recorded, and which specific visual technology is preferred depending on different AAL contexts.

Thereby, the specific underlying research questions were the following:

  • RQ1: (To what extent) Do the affective perceptions of a human and a technological privacy invasion differ?

  • RQ2: Which concerns are relevant considering a human and a technological privacy invasion?

  • RQ3: (To what extent) Does the privacy context (general vs. caring in older age) impact the privacy preferences in terms of their acceptance of recording activities of daily living?

  • RQ4: (To what extent) Does the privacy context (protecting privacy vs. caring in older age) impact the selection of a specific technology for recording activities of daily living?

3 Empirical approach

This section describes the underlying empirical approach of the present study, starting with the design of the online survey and the applied procedures of data analysis. Further, the characteristics of the sample are presented.

3.1 Design of the online survey

Prior to the quantitative investigations presented in this study, qualitative preceding research identified existing mental models of privacy, relevant concerns, and requirements in the context of privacy as well as decisive invasions of privacy (affective perceptions of privacy violations, recordings of activities, persons being allowed to invade privacy [47, 48]). These qualitative insights and results represented the basis for the conceptualization of a subsequent quantitative study in terms of an online survey. This quantitative online survey was applied to answer and analyze the described research questions (see Fig. 1), aiming for an investigation of the perceptions of different privacy invasions as well as privacy preferences in using visual ambient assisted technology for different application contexts. Initially, the participants were welcomed and informed about the topic of the online survey.

Fig. 1
figure 1

Schematic overview of the online survey parts

Within the first part of the survey, the participants indicated their demographic data, i.e., age, gender, educational level, and living situation. Here, the participants also responded to questions about their general state of health, presence/absence of chronic disease(s), the necessity of care in their everyday life and their previous experiences in care.

The second part of the survey focused on the participants’ perception of diverse invasions of their own privacy differing between a human and a technological invasion. For this purpose, the participants assessed 11 adjective pairs of a semantic differential (see Figs. 1, 2) based on a six-point Likert scale (e.g., 1 = happy; 6 = sad) for a human and a technological way of privacy invasion. The adjective pairs were taken from the preceding qualitative assessment mentioned above. Further, the participants selected relevant concerns for both types of privacy invasion as well. For each type of privacy invasion, they selected the most relevant concerns out of eight items.

Fig. 2
figure 2

Schematic results of a semantic differential comparing feelings of human and technological invasions of privacy

Besides rather generic evaluations of the perceptions of privacy invasions, this study aimed at the investigation of more specific privacy perceptions in different contexts. Therefore, the last part of the survey asked the participants to empathize with the situation that visual ambient assisted technology is installed in their home environment. Subsequently, the participants evaluated their acceptance or rejection in terms of their comfortableness of recording 16 different activities of daily living (see Fig. 3) based on six-point Likert scales (min = 1: “very uncomfortable”; max = 6: “very comfortable”). In a first step, the participants should perform the evaluation in a General scenario empathizing the usage of visual ambient assisted technology within a normal daily routine at the own home. In a second step, the participants should perform the same evaluations within a Caring in Older Age scenario: Here, the participants should imagine that they are at an older age (approx. 80 years), physically limited in everyday life due to some illnesses, depend on a little support, and live alone at home.

Fig. 3
figure 3

Comfortableness of using VAAL comparing a “General” and a “Caring in Older Age” usage context

Beyond that, the participants were asked to select their preferred types of visual ambient assisted technology for two different scenarios differing between (1) a conventional camera, (2) a depth camera, or (3) none of the camera types. Initially, the participants were asked to select their preferred type of visual ambient assisted technology which they think is best suited for Protecting their own Privacy. Then, the participants performed the same selection but this time they should choose the type they think is best suited to monitor an older person in need of care (Caring in Older Age).

At the end of the survey, the participants had the opportunity to leave comments and to provide feedback to the content or the online survey itself on an optional basis.

3.2 Data Analysis

Besides descriptive statistics (means (M), standard deviations (SD), and relative frequencies), T Tests for paired samples were used to investigate differences between human and technological invasions of privacy as well as influences of different contexts on privacy perceptions and preferences. Thereby, the T-statistic is reported as a calculated test characteristic and Cohen’s dz is reported for effect sizes. The level of significance was set at p < 0.05; and therefore, values above the significance level of p > 0.05 were interpreted as not significant (n.s.).

3.3 Description of participants

Overall, N = 139 German individuals participated in the study and filled out the online survey nearly completely. The data collection took place in Germany in November and December 2021. A rather broad sample was targeted ranging from young participants to older adults to cover an extensive portion of future potential users of AAL technology. Indeed, everyone might get involved in care need. Rather young participants might be involved in the decision of introducing AAL technology for a close older relative and older participants might face this decision for themselves in the near future. Participants were invited to take part in the online survey via link distributed within the social networks of the authors and online forums.

The mean age of the participants was rather young (M = 31.20, SD = 14.71, min = 17, max = 69), and the sample consisted of a higher proportion of female (66.20%, n = 92) compared to male (33.1%, n = 46) participants (0.70% (n = 1) reported a diverse gender). The majority of the participants reported a high educational level in terms of a university degree (26.60%, n = 37), university entrance qualification (54.00%, n = 75) or a PhD (0.70%, n = 1). Only 18.70% (n = 26) of the participants reported lower educational levels, such as secondary or elementary school certificates. Asked for their living situation, most of the participants (64.00%, n = 89) indicated to live together with several people, while smaller proportions reported to live together with another person (24.50%, n = 34) or alone (11.50%, n = 16). Regarding health-related characteristics, a great majority of the participants perceived their health status as very good (30.20%, n = 42), good (52.50%, n = 73), or rather good (15.8%, n = 22). In contrast, only 1.40% (n = 2) of the participants evaluated their health status as rather bad or bad. In line with the positive health perception, only a fifth of the participants (20.10%, n = 28) reported to suffer from a chronic disease, while none of the participants indicated to be in need of assistance and care in their everyday life. Further, only small proportions of the participants indicated to be professionally (12.90%, n = 18) or privately (20.1%, n = 28) experienced in caring and supporting people in need of care.

4 Results

In the following, the results of the empirical study are presented, starting with a comparison of human and technological invasions of privacy as well as the relevance of related concerns (see 4.1). Subsequently, it is shown to what extent privacy needs depend on different usage contexts (see 4.2).

4.1 Comparing human and technological invasions of privacy (RQ1 & RQ2)

In a first step (RQ1), the participants were asked to evaluate their feelings when their privacy is violated by a human or by a technology. For this purpose, a semantic differential with 11 pairs of adjectives was used and the schematic results are presented in Fig. 2. Not surprisingly, all evaluations tend more toward the negative poles of the adjective pairs (right side, mean values > 3.5) independent from a human or a technological invasion of privacy. All relevant statistical information is presented in Table 1. There, it can be seen that invasions of privacy independent from a human or a technological way were associated to awake feelings related to the negative poles “sad,” “disappointed,” “deceived,” “unpleasant,” “feeling ashamed,” “angry,” and “worst thing happened to me in my life.” These evaluations did not differ significantly depending on a human or a technological invasion of privacy (n.s.). In contrast, a technological invasion of privacy was perceived significantly more negative regarding the pole “devastating” (p < 0.05) and even stronger regarding the poles “feeling at the mercy of others,” “powerless,” and “defenseless” compared to a human invasion of privacy. These last three differences were at a high level of significance (p < 0.001) and possessed low to medium effect sizes (dz = 0.42–0.57).

Table 1 Descriptive and inference statistics of the semantic differential related to invasions of privacy

In addition to the assessments of feelings, the participants were asked to select concerns they associate with the situation of a human invasion of their own privacy (RQ2). They were allowed to select multiple answers to enable a broad overview of all relevant concerns. Table 2 shows the respective results in terms of absolute selections and relative proportions related to the whole sample of participants. The majority of the participants (78.42%) showed concerns that “…people will spread the information.” Further, concerns regarding “… being judged” (53.96%), “…no longer feeling protected at that particular place” (41.01%), and “… many people will know about it” (37.41%) represented highly relevant concerns related to the situation of a human invasion of the own privacy. In a direct comparison, concerns “that what happened will happen again” (19.42%) and “… about being pitied” (10.79%) were selected clearly less often. “I am worried about not being able to mentally process what has happened” (6.47%) represented the least selected and thus least relevant concern regarding a human invasion of privacy. The low selection rate of the statement “I am not worried about this” (10.79%) indicates a high relevance of privacy perceptions regarding potential human invasions.

Table 2 Mentioned concerns regarding a human invasion of privacy

In line with human invasions of privacy, the participants were also asked to select concerns they associate with the situation of a technological invasion of their own privacy (see Table 3). Here, it was also allowed to select multiple answers in order to allow a broad overview of all relevant concerns. Overall, the concern “… that my data will be misused” (89.93%) was selected by the great majority of the participants and thus represented the most relevant concern related to a technological invasion of the participants’ own privacy. Further, the concern “… about which people can see my data” (69.06%) was selected by more than two thirds of the participants. Compared to that, the concerns “… an uncomfortable feeling” (39.57%) and “I feel like I am transparent to many people” (30.22%) were selected less often, but still represented relevant concerns of a technological invasion of privacy. The selection of the statement “I feel less embarrassed because technology cannot judge and evaluate me” (28.06%) indicates differences between a technological and a human invasion of privacy. The low selection rates of statements related to little or no worries (last three statements in Table 3) show the high relevance of privacy perceptions and related concerns regarding technological invasions.

Table 3 Mentioned concerns regarding a technological invasion of privacy

4.2 Investigating privacy needs in different contexts (RQ3 & RQ4)

In order to concretize privacy perceptions and preferences, the participants were asked to empathize with different scenarios and to imagine using visual ambient assisted living technologies in their home environment. For two different application contexts, the participants should assess to what extent they feel comfortable that visual ambient assisted living technology records different activities in their everyday life (RQ3). In a first step, the evaluations should be performed within a general usage context as part of a normal daily routine within the own home environment (General). In a second step, the participants should empathize with the scenario that the visual ambient assisted technology is used to monitor them as an older person in need of care (Caring in Older Age, see Sect. 3.1).

Figure 3 illustrates the evaluations referring to both usage contexts; whereas, Table 4 contains all relevant descriptive and inference statistical information. From both, the Figure and the Table, it can be seen that a visual recording of all single activities was rejected to a different extent for the general usage context (indicated by means < 3.5). Instead, recording activities like “cooking,” “working,” “cleaning,” “tidying up,” “sports activity,” “leisure activities,” “doing laundry,” “eating,” “playing,” and “brushing teeth” was evaluated significantly more positive revealing accepting values (> 3.5) for the context Caring in Older Age. Further, the strongly rejected activities for the General usage context (“sleeping,” “washing,” “changing clothes,” “showering,” and “going to the toilet”) were evaluated less negative, but still resulting in rejecting evaluations for the Caring in Older Age context. All these differences were at a high level of significance (p < 0.001) and possessed medium to high effect sizes (dz = 0.66–1.11).

Table 4 Descriptive and inference statistics of the acceptance of context-related usage of VAAL technology

As a last aspect (RQ4), the participants should select which specific visual ambient assisted technology they would prefer to be used in two different contexts: (a) for protecting privacy and (b) for caring in older age. As introduced in Sect. 3.1, the participants should decide between the options of “conventional (color image) camera,” “depth camera,” or “none of the camera types.” The respective relative selections are presented in Fig. 4 and the results of a variance analysis with repeated measures revealed distinct selection patterns (F(1,138) = 74.88; p < 0.01; η2 = 0.35).

Fig. 4
figure 4

Technology selection comparing two contexts: protecting privacy and caring in older age

Starting with the context of Protecting Privacy, almost all participants decided to prefer using the depth camera (92.10%, n = 128), while only small proportions of the participants chose the conventional (color image) camera (2.90%, n = 4) or none of the camera types (5.00%, n = 7). Within the context Caring in Older Age, the selection patterns changed fundamentally: Here, most of the participants (60.40%, n = 84) preferred using the conventional (color image) camera, followed by more than one third of the participants (36.70%, n = 51) selecting the depth camera. In line with the evaluation in the context of Protecting Privacy, the option none of the camera types was selected by very few participants (2.90%, n = 4).

5 Discussion

This study provided insights into a first quantification of specific privacy perceptions, concerns, and requirements when visual AAL technologies are used. Thereby, distinct differences in affective evaluations and concerns regarding human and technological privacy invasions were identified. Beyond that, the results revealed a context-dependent evaluation of visual recordings of daily activities which are additionally affected by the sensitivity of the activities. In line with this, the selection of a specific visual AAL technology was also impacted by the usage context. In the following, we discuss the key insights in more detail and derive recommendations and guidelines to consider the future users’ requirements in the development and design of visual AAL technology.

5.1 Key insights related to the research questions

Addressing RQ1 and thus the comparison of a human and technological invasion of privacy, the results did not show significant differences regarding general feelings (sadness, satisfaction, respect, inspiration, pleasantness, relaxation, self-confidence, and anxiety). In contrast, we identified differences concerning emotions dealing with protection, control, and safety: Here, a technological invasion of privacy is perceived more negatively indicated by stronger associations with the negative poles of the adjective pairs (compared to a human invasion of privacy). Thereby, this study’s findings represent a deepening and extending of the results of Ray et al. [46]: The assumed differences between human and technological privacy invasions are quantified and can be verified in this study; however, the identified differences can be predominantly attributed to the perceptions of protection, safety, and control, but not to other rather generic feelings and emotions. In general, privacy is seen—among others—as an assertion of control [49] and this aspect seems to be lacking significantly more when dealing with technological devices. Apparently, participants felt to have more available behavioral mechanisms at hand to be used to defend and control the invasion coming from a human privacy intruder compared to technology with seemingly unpredictable functions and possibilities of gathering, storing and sharing information.

Considering perceived concerns comparing a human and a technological invasion of privacy (RQ2), different focal points become evident. The most relevant concern of a human invasion of privacy related to the aspect that people will spread the information. Other relevant concerns were fears about being judged, no longer feeling protected and that many people will know personal information. Overall, feelings dealing with protection and losing control were decisive for human invasions of privacy. Taking the technological invasion of privacy into account, the participants selected more concerns compared to a human invasion of privacy, and the concern that data will be misused represented the most selected and most relevant aspect. Further, also concerns about which people can see the data, uncomfortable feelings and feared transparency were frequently selected. Here, data security, but also feelings related to an undesirable transparency and discomfort are decisive. However, it seems, that concerns regarding privacy invasions have similar roots and that concerns regarding technological invasion are an extension of the concerns raised in cases of human invasion. Indeed, spreading information and fears of being judged can relate to the technological fears of misusing information, in the sense of massively and inappropriately spreading and accessing of personal information by technological means which may evoke the uncomfortable feeling of potentially being (mis)judged by strangers.

Further, our results identified that privacy perceptions in terms of the comfortableness of recording daily activities were clearly impacted by the context of technology usage (RQ3). Considering similarities of both contexts, recording of sensitive activities (such as changing clothes, washing, going to the toilet, or showering) was rejected for both contexts, but to a different extent. A visual recording of typical daily activities (e.g., cooking, eating) and social activities (e.g., leisure activities, sports) was in tendency more accepted in terms of comfortableness. In some way, these results fit to and extend the results of Himmel and Ziefle [16], who found that technology usage is not accepted in sensitive and private rooms, such as the bathroom, where the mentioned sensitive activities mostly take place. However, our results also showed that the acceptance evaluations differed depending on the context of technology usage. For a general usage context of AAL technology, the visual recording was rejected for every kind of daily activity; however, recordings of typical daily and social activities are slightly rejected, while sensitive activities (e.g., showering, going to the toilet) are strongly rejected. Taking the caring in older age context into account, the visual recording of daily and social activities was accepted, while recordings of sensitive activities were still rejected. In line with previous research results, the results suggest that the participants weigh the advantages (i.e., increase in safety) against the fear of privacy invasion [50]; in case of daily and social activities—which are presumably less sensitive and privacy intrusive—the benefits of increased safety superimpose the privacy concerns. In case of very sensitive activities, it is exactly the other way round so that privacy concerns superimpose the benefits.

Finally, we also focused on privacy preferences regarding the usage of different visual technologies in different contexts (RQ4). The results indicate distinct selection patterns of the participants: in the context of protecting the own privacy, almost all participants selected the depth camera to be the preferred visual AAL technology. Thus, to protect privacy, using a depth camera is perceived as an optimal option when visual AAL technology is employed. Changing the context to caring in older age led to a less clear decision pattern: The majority preferred a conventional (color image) camera, while still a third of the participants preferred the depth camera. The results suggest the assumption that again a trade-off has been made between the advantages and the privacy concerns (e.g., [50]). Here, the feedback of our participants in open comment fields showed that a majority of the participants chose the conventional color image camera with the expectation that emergency situations can be recognized more precisely and efficiently: In that case, the benefits of quick help and increase security were more important than existing privacy concerns; for the participants selecting the depth camera privacy was still more relevant. It should be considered that the participants evaluated the camera types purely based on the visualization characteristics and not in terms of technical details. Of course, it could be that specific information about, and the differentiation of technical details affect the participants’ assessment. Therefore, at this point a future research recommendation can already be anticipated. This means that future research should focus more intensively on specific comparisons of visualization and data elaboration modalities of visual AAL technologies {e.g., [51,52,53]) as well as various other technical applications of AAL (e.g., [54, 55]) differentiating in (in)visibility and analyzability of recorded data based on applying AI- or multi-sensor-based approaches.

5.2 Derived recommendations and guidelines

The results showed that protection, safety, and control are of major importance when people think about their privacy perceptions and technology usage. For the development of AAL technology, it is therefore recommended to inform and communicate these aspects transparently: Future users should receive information about how their data is protected, which specific data security measures are defined and realized, and who has access to the recorded data at which time. In addition, it should be clearly regulated by whom, when, and under which conditions the specific technology is controlled (e.g., turning on and off). These aspects should be communicated transparently and comprehensively to enable more comfortable feelings in using the technology, to avoid the fear of too much transparency of individual information, and to facilitate trust toward developers and manufacturers.

This study revealed that the context in terms of the purpose of technology usage impacts the acceptance of a visual recording of daily activities fundamentally, resulting in negative evaluations of using visual AAL technology in a general everyday life scenario. Here, it is recommended to not aim at an “all-in-one” visual AAL technology solution covering diverse application areas. From an acceptance perspective, it is more useful to tailor a technological approach to a specific application explaining the purpose and respective advantages of using the technology in that case, e.g., caring for an older person in need of care. Beyond that, it has to be considered that visual AAL technology is not accepted to record all activities of daily living likewise. Here, individual profiles could be a solution enabling the involvement of the individual user in deciding which activities may be recorded. In line with this, the selection of specific visual technologies is highly individual and context-dependent. Again, the purpose of technology usage is decisive and should be focused on in the decision process: e.g., if the main goal of technology usage lies on the protection of the individual privacy, the usage of depth cameras can be recommended. For specific contexts, e.g., caring for an older person in need of care, there is no distinct recommendation. Here, it should be discussed and decided individually which technical solution fits the requirements best. In addition, it has to be considered that only two visual AAL technologies (and a none-option) were provided in this study: Here, future approaches should also include and investigate other visual approaches (e.g., other camera types or different filters of visualizations) in order to identify ideal approaches to cover and fulfill future users’ privacy requirements. These requirements should be studied in collaboration with technological experts to account for technological feasibility and technological options which are realistic.

5.3 Limitations and future work

This study enabled a first quantification of privacy perceptions, concerns, and requirements considering the usage of visual AAL technologies. However, there are some limitations regarding our methodological procedure which should be considered for future research.

First, our sample was limited to the German population, comparably small, and rather young containing a higher proportion of female compared to male participants. As the aim of this study was to investigate potential user groups and generations of tomorrow, the results revealed useful insights enabling to derive recommendations for ongoing technology development and design addressing the requirements and needs of these future user generations. This was indeed a timely and convenient decision as the development of visual devices for AAL are also still in development and not yet widely implemented in Germany. However, a larger and more balanced sample (e.g., regarding age, gender, and technology expertise) should be investigated in future studies as this would enable to investigate the influences of individual user characteristics, which was not aimed at in this first quantitative study. Beyond that, a specific focus on old and frail users who would directly profit from using visual AAL technologies is absolutely desirable and should be realized in future research. For that purpose, it would be necessary to move away from online surveys and predominantly use paper-based questionnaires in order to address older, frail, and less tech-savvy participants.

Our results suggest that there are different categories of activities (e.g., normal daily activities: eating, cooking; social activities: leisure activities, sports; sensitive activities: showering, going to toilet). Besides the latter categories, more specific care—related activities (e.g., feeding, incontinence insoles exchange) as well as unintended activities (e.g., accidents, interruptions) become relevant when visual AAL is used. In future studies, these activity categories including the specific care related and unintended activities, should be compared and analyzed in more detail regarding privacy requirements, considering aspects such as the duration and location of data storage, data access, or willingness to share data. Direct evaluations and comparisons of the activity categories would enable to derive even more concrete design recommendations and guidelines.