3.2.1 Method.

The survey was built in Unipark (https://www.unipark.com/en/) and is structured in four parts: navigation behaviour description, navigation mode preference, navigation frequency and demographics. Two attention checks were built into the survey to verify that participants gave valid answers. The full questionnaire can be found in Appendix.

The data collected through the survey were analysed using descriptive statistics and binomial tests to examine preferences in navigation modes across various scenarios. While the initial sample was representative of the U.S. population, some participants were excluded from the final analysis based on specific criteria to ensure the relevance of the data to pedestrian navigation. Participants who never used navigation apps (37), indicated that they didn’t see a difference between the two navigation modes (10), never walked as a mode of transportation (40), or never used navigation apps while walking (48) were excluded from the final sample. These exclusions were necessary to focus the analysis on participants with relevant navigation experience and understanding. After applying these exclusion criteria, the final sample consisted of 222 participants (75% of the original sample). This final sample was used for all subsequent analyses.

Demographics: The survey was completed by 222 participants: 101 female, 115 male, 5 diverse, and 1 preferred not to state their gender. The mean age was 43.9 years, ranging from 18 to 74 years (). Regarding formal education, 49.6% had a bachelor’s degree, followed by a high school diploma (29.7%) and a master’s degree (16.7%). 3.6% had a doctoral degree, and only one person had no formal education (0.45%). With 55.9%, the leading operating system used by the participants is iOS by Apple. The use of an Android system was reported by 44.1% of the participants. Regarding the navigation app, 74.8% of the participants reported using Google Apps. Apple Maps is used by 20.7%, and “other” apps by 4.5% of the respondents. A majority of the participants (56.3%) self-reported their proficiency using navigation apps as advanced. Intermediate was the second most given answer (22.5%), while expert was the third most picked option (20.3%). Only 2 participants (0.9%) self-reported as beginners.

To increase ecological validity, participants were asked to recall their most recent navigation experience before answering preference questions, grounding their responses in actual use cases rather than hypothetical scenarios.

Frequency of Transportation Modes: Table 1 presents the responses of participants regarding how frequently they use different modes of transportation and how often they use navigation apps while moving. The “never” category for walking and use of navigation apps while walking shows 0% due to the data preparation that was mentioned before. The results show a clear contrast between the two transportation modes. While driving is a daily activity, walking isn’t as common. Additionally, navigation apps are used less frequently while walking, with most individuals relying on them only occasionally. In contrast, driving sees a higher dependency on these apps, with a majority of respondents using them regularly.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Study 1: Frequency of transportation mode usage and navigation app use.

https://doi.org/10.1371/journal.pone.0340711.t001

3.2.2 Results.

In this subsection, we present the results of the survey, which include the 222 participants remaining after the previous data preparation step.

Mode Preference: 43.7% (97 participants) of the respondents chose the RP as their preferred navigation mode, and 56.3% (125 participants) selected TBT. A binomial test showed no statistically significant () for the observed preference. For Apple Maps, the preference for TBT is higher than across the full sample, with 69.6%, while for Google Maps, it is slightly lower, with 53.0% preferring TBT navigation. To examine if the preferred navigation mode differs significantly between demographic groups and navigation scenarios, we conducted a series of Chi-Square and binomial tests. We tested for differences based on age groups, gender, user operating system, and type of navigation app (Apple Maps vs. Google Maps). None of these factors showed statistically significant differences in navigation mode preference.

Mode Preference in Scenarios: When navigating in familiar areas, a majority of participants (169 / 76.1%) preferred the RP mode, while a minority (53 / 23.9%) chose the TBT mode. A binomial test found a significant preference for RP mode in familiar areas (). In unfamiliar areas, the preference shifted, with 156 participants (70.3%) preferring the TBT mode and only 66 people (29.7%) choosing the RP mode. A binomial test found a significant preference for TBT mode in unfamiliar areas (). These results suggest that the familiarity of the area has a strong influence on users’ navigation mode preferences, with a clear tendency towards RP in familiar areas and TBT in unfamiliar areas.

For longer routes than usual (based on the participants’ self-assessment), 116 participants (52.3%) preferred the TBT mode, whereas 105 participants (47.3%) selected the RP mode. This difference was not statistically significant (binomial test, ). Similarly, when navigating shorter routes than usual (based on the participants’ self-assessment), 117 respondents (52.7%) preferred the RP mode, compared to 104 participants (46.8%) who chose the TBT mode, again with no significant difference (binomial test, ). For the total of 222 participants using pedestrian navigation apps, 14.4% prefer exclusively RP and 12.6% prefer exclusively TBT navigation. The remaining 73.0% of participants changed their preferred navigation mode based on different scenarios.

Features for Route Preview: Lastly, we asked participants to report features that would improve upon the current implementations of the RP mode. We performed a frequency analysis on the answers. Landmarks were by far the most desired features by the respondents, while the other four top features were mentioned only a few times.

1. Landmarks (mentioned 27 times)
2. Street View (mentioned 8 times)
3. Points of Interest (mentioned 7 times)
4. Customisable (mentioned 5 times)
5. Automatic Map rotation (mentioned 3 times)

Summary: The survey revealed that a considerable share of participants actually preferred the RP mode over the TBT mode. While this preference suggests potential areas for enhancing current RP navigation designs and justifies further exploration, our subsequent study first examined whether there are measurable differences in performance between the two navigation modes.

3.3.1 Method.

This section outlines the methodology employed in our study, detailing our study design, metrics, data preparation and further statistical foundation.

Study Design & Procedure: We recruited 218 students who had recently enrolled at Monash University in Melbourne, Australia, as our participants. All participants took part in the experiment within three weeks to ensure similar outdoor and infrastructural conditions between the 21st of March 2023 and the 11th of April 2023. We followed a citizen-science approach [49,50], where the participants would perform the experiment on one another: one being the navigator using the app, the other monitoring the navigator (i.e., the “experimenter”). To ensure the reliability of the experiment, we conducted a pilot to verify that the instructions were easy to follow. We also prepared video tutorials for each task and made both a briefing session and assistance available to anyone unsure about the requirements. This is in line with best practice for citizen science experiments [51].

Our participants were assigned to one of two conditions: TBT navigation or RP navigation. Participants were randomly assigned one of 20 predefined routes located near the campus. The 20 routes were selected using the Google Maps API, with the route length kept between 1400 and 1600 meters, featuring 9–11 manoeuvres—a common setup for outdoor navigation studies [13] to ensure consistency. A manoeuvre is defined as a navigation action to take for the current step (e.g., turn left, merge, straight, etc.). The experiment was conducted using the current version of Google Maps (version 11.69.0401 for Android and version 6.57.0 for iOS), which was used by all participants. A GPS recording of the route navigated was taken in the background, using OSMAnd. Whilst travelling the route, experimenters were instructed to walk within 10 meters of the participant and monitor them closely. They noted down any navigation error, where a participant accidentally left the route, as well as the number of times participants looked at their smartphones. Immediately after completing the journey, the navigator was asked to complete a route sketching task [52,53], where they would sketch the route as they remembered it. The participants also completed a demographic survey, whose results are summarised in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Study 2: Participant demographics, preferences, and age statistics.

https://doi.org/10.1371/journal.pone.0340711.t002

We recognise that TBT navigation was not originally developed to support spatial learning, as it was designed primarily for in-car navigation. Previous research indicates that TBT requires relatively low cognitive engagement compared with RP, which involves more active processing of spatial information and may therefore facilitate stronger spatial learning outcomes. Nevertheless, our comparison reflects realistic navigation conditions, in which users are often required to recall routes regardless of the navigation mode employed. To ensure a comprehensive and balanced evaluation, several methodological controls were implemented.

First, we selected performance metrics capturing both navigation efficiency, including Navigation Errors and Phone Glances, and spatial learning, assessed through the Route Sketching Task using the metrics Correctly Indicated Turns and Correctly Indicated Directional Changes. Second, we recruited participants with minimal familiarity with the study area, focusing on students who had recently enrolled at Monash University. Third, a pre-study familiarity assessment (summarised in Table 2) indicated that 44.11% of participants rated themselves as unfamiliar or very unfamiliar with the routes and over 69.2% including the neutral participants. Finally, we controlled for familiarity effects through multiple linear regression analysis (Table 3), which confirmed that increased familiarity was significantly and negatively associated with the number of phone glances, providing a validity check for our data. Although the Route Sketching Task may inherently favour RP due to its interface encouraging higher cognitive engagement and spatial encoding, it also reflects a realistic situation in which users must navigate without digital assistance or communicate routes to others. This measure is therefore critical for evaluating spatial knowledge acquisition, which represents a central advantage of the RP mode.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Study 2: OLS regression results. The regression analysis tested the significant influence of confounding variables in our study. This analysis included gender, age, familiarity with the experiment environment, the frequency of how regularly participants use mobile navigation apps, and their personal preference for TBT or RP navigation. Statistically significant results are marked with asterisks: .

https://doi.org/10.1371/journal.pone.0340711.t003

Metrics: Our primary goal was to understand how navigation performance, attention during navigation, and spatial learning differ between TBT navigation and RP navigation. Given mixed findings in the literature in terms of the performance of those two modes, we did not setup specific hypotheses. However, we define four key metrics that will serve as performance indicators common for navigation experiments [12,52,54–56]:

• Navigation Errors: This metric quantifies the number of mistakes made by the participant during the navigation task. An error is defined as any deviation from the predefined route. A lower number of errors indicates higher navigation performance.
• Phone Glances: This metric measures the frequency with which participants looked at their smartphones during the navigation task. Frequent glances may indicate greater reliance on the navigation system and potentially reduced awareness of the surrounding environment.
• Correctly Indicated Turns: This metric gauges the accuracy of the participants’ route sketching task by counting the number of turns (left, right, etc.) correctly indicated on their sketched maps. A higher number suggests better spatial learning and memory of the route.
• Correctly Indicated Directional Changes: Similar to “Correctly Indicated Turns,” this metric also evaluates the accuracy of the participants’ route sketching task. However, it focuses on capturing broader directional changes, such as merging or going straight, as opposed to just turns. Like the previous metric, a higher count suggests better spatial understanding and memory of the route.

These metrics collectively provide insights into the navigation performance, attentional engagement, and spatial learning outcomes of participants, thereby enabling a comparison of the effectiveness of TBT navigation and RP navigation.

Data Preparation: Before our evaluation, we used a two-step approach to clean and prepare our data. In the first step, we removed entries identified as faulty data due to participants not having thoroughly recorded or submitted their data. For this, we used indicators such as map sketch results of 1000% or over 100 navigation errors. For each such value that we noticed, we reviewed any comments made by the experimenter that suggested there had been an execution error with the experiment. Only if these two indicators matched were we able to remove the participants from the dataset. Furthermore, we removed entries that stated they had to use a custom route instead of the provided ones, due to participating remotely, which resulted in routes with different lengths and numbers of manoeuvres outside the defined range for our study. After cleaning the data, 208 out of 218 participants remained in the dataset. In a second step, we identified outliers and removed those with a Z-score above 3 to exclude extreme outliers for statistical significance testing. After removing these outliers, 195 participants remained in the final dataset. All statistical tests in our analysis were conducted on both the raw and cleaned datasets, yielding consistent results.

Statistical Methods: To test our data for normality, we applied a Shapiro test to our data for each condition and metric, which resulted in a for all results. Thus, we assume that our data is not normally distributed. Since our data are also unpaired, the Mann–Whitney U test was used to compare results between the two conditions, unless otherwise specified. Effect sizes are calculated using the rank-biserial correlation (denoted with r in the results). Furthermore, we used multiple linear regression to understand the relationship between the predictor variable (the experiment condition), our dependent variables, and other covariates.

Power Analysis: We used the G*Power software to conduct a power analysis for our study. According to Robertson and Kaptein [57], effect sizes for non-parametric tests like the Mann-Whitney U test typically range from 0.1 for a small effect, 0.3 for a medium effect, and 0.5 for a large effect. Given our study design and common assumption for power and significance, detecting a large effect would require 134 participants, a medium effect would require 368, and a small effect would require 3290.

With 195 participants, our study is sufficiently powered to detect a large effect. We chose to focus on detecting a large effect because TBT navigation is the default and far more prevalent mode in mobile navigation apps compared to RP. Thus, only a large difference in effectiveness between these modes would warrant the current design choice of TBT being the default. A small or medium effect, while potentially statistically significant, would not be sufficient to advocate for such a decision; it would rather argue for the co-existence of both modes to cater to user preferences.

Balanced Random Assignment to Conditions: Our study utilised random assignment to distribute participants across two distinct navigation conditions: RP and TBT. To validate the effectiveness of this randomisation process, we examined key demographic and behavioural variables. The mean age in the RP condition was approximately years, while it was years in the TBT condition (). Both conditions displayed a balanced gender distribution between each other with around females and males. of participants in the RP condition and in the TBT condition preferred the navigation method they were assigned to. Although this could be influenced by the timing of the post-experiment demographic questionnaire, it is noteworthy that about of participants in each condition preferred the opposite navigation method. This suggests that any potential bias introduced is likely balanced across both conditions. Overall, these findings affirm the success of the random assignment to conditions.

3.3.2 Results.

After cleaning the data and removing extreme outliers, a total of 195 participants (out of the initial 218) contributed valid data to the study. Their demographic information and participant preferences are summarised in Table 2. In the following section, we present our results of the study.

Navigation Errors: During the navigation task, the experimenter noted any navigation errors made by the participant. On average, participants made more errors in the TBT condition (, , ) compared to the RP condition (, , ). Even though this was not statistically significant (, , ), there is a trend towards TBT navigation resulting in more errors (see Fig 1). In a OLS regression model, none of the variables were found to be statistically significant control variables for navigation errors. For detailed coefficients and -values, please refer to Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Study 2: Mean values of key navigation metrics between the two conditions: turn-by-turn navigation and route preview navigation.

https://doi.org/10.1371/journal.pone.0340711.g001

Attention During Navigation: The experimenter noted down the number of times the participant looked at their phone. On average, participants looked at their phone more often in the RP condition (, , ) than in the TBT condition (, , ). While this was not statistically significant (, , ), we see an indication that RP navigation results in more attention on the mobile navigation device (see Fig 1). This observation, however, fits the notion of related work that RP navigation might demand more elaborate processing of spatial information [10] and thus lead to more increased phone glances. An OLS regression model for the phone glances (looks at the navigation device) found that the variable ‘Familiarity’ was statistically significant () and negatively correlated with the number of glances at the phone. This suggests that participants who are more familiar with the area tend to look at the navigation device less frequently, which is a logical finding and thus serves as a sanity check for our data. Table 3 shows a complete list of coefficients and -values. A subsequent correlation analysis revealed a statistically significant negative correlation between ‘Familiarity’ and the number of ‘Phone Glances’ at the navigation device (, ). This suggests that as participants’ familiarity with the area increased, they tended to look at the navigation device less frequently.

Route Sketching Task: After completing the navigation task, participants were asked to sketch the route they had just walked on a piece of paper. The evaluation of the sketch consisted of two parts: the percentage of correctly indicated turns and the percentage of correctly indicated directional changes.

Correctly Indicated Turns: Participants, on average, correctly indicated more turns in the RP condition (, , ) than in the TBT condition (, , ). There was no statistically significant difference between the two (, , ), but the trend points towards RP resulting in better route sketching abilities. The regression model for indicated turns found no variables to be significant control variables for the correct number of turns indicated. For further details, please see Table 3.

Correctly Indicated Directional Changes: For the correctly indicated directional changes, participants also correctly indicated more turns in the RP condition (, , ) than in the TBT condition (, , ).

An OLS regression model found that the variable ‘Frequency of Using Navigation Apps’ was statistically significant () and positively correlated with the percentage of correctly indicated directional changes. This suggests that participants who frequently use navigation apps tend to make more accurate directional changes. For additional coefficients and -values, refer to Table 3. A subsequent correlation analysis revealed a statistically significant positive correlation between ‘How often users use mobile navigation apps’ and the percentage of ‘Correctly Indicated Directional Changes’ (, ). This suggests that as participants’ amount of general mobile app usage increases, their ability to correctly indicate turns in our route sketching task improves.

Analysis of Navigation Method Preference: We investigated the potential confounding factor of whether participants used or did not use their preferred navigation method. To assess the impact of this factor, we conducted Mann-Whitney U tests for our four dependent variables: number of errors made, number of times the participant looked at the navigation device, the percentage of correctly indicated turns, and the percentage of correct directional changes. Our analyses revealed no statistically significant differences between the group of participants who used their preferred navigation method and those who did not, across all four variables. This suggests that the preference for a particular navigation method did not significantly affect the outcomes in our study.

Post-hoc Power Analysis: Our analysis across all dependent variables revealed small effect sizes, which have critical implications for the interpretation of our results. For instance, in the case of navigation errors, a post-hoc power analysis demonstrated that with an effect size of 0.06 and our current sample size, the achieved power was only . To detect an effect of this magnitude as significant, a substantially larger sample size would be required.

Interestingly, the TBT navigation condition actually performed worse than the RP condition, a finding that is surprising based on current assumptions in the field. However, it is crucial to note that the observed effect sizes are not only statistically insignificant but also practically negligible. Even if statistical significance could be achieved with a larger sample, the practical impact of these effect sizes would be minimal. This reinforces the idea that the prevailing implementation of TBT navigation may need to be re-evaluated, especially given its counterintuitive, poorer performance in our study.

3.4.1 Method.

We used the design studio method to conduct our workshop [58]. The goal of the design studio was to develop concrete interface designs for features that participants would like to see in combination with the RP navigation mode. The workshop was organised into several phases. It began with an introduction by the facilitators, who were two of the paper’s authors. This introduction included an overview of the workshop schedule and a brief presentation of the research, summarising the related work discussed in Section 2. During the interactive parts, participants generated ideas, developed them further, reflected on their relevance, and finally prioritised them. This process resulted in the identification of the most crucial features for enhancing the user interface. The workshop, which lasted three hours, took place on June 12, 2024. We recruited five participants [59,60], all of whom signed a consent form. The workshop was audio- and video-recorded.

Demographics: Participants were, on average, 26.4 years old (). The group consisted of two female and three male participants. Their educational backgrounds ranged from high school diplomas to doctoral degrees, offering a diverse range of perspectives. Three participants used Apple’s iOS, while two used an Android system.

3.4.2 Results.

This section presents the exploratory findings of the workshop. The following subsections outline the design implications for Landmarks, ETA Updates, and Map Orientation, beginning with the highest-priority topic identified by the workshop participants.

Landmarks: Landmarks emerged as a crucial feature in our design studio workshop, with participants unanimously endorsing their integration into the RP mode. Defined as important spatial features used as reference points [61], landmarks play a vital role in everyday navigation strategies. Participants recounted using landmarks to orient themselves in familiar cities and to correct navigational errors:

“When I’m looking for a restaurant in the city that I haven’t been to before but know the city relatively well, I approach it differently. In such a case, I would simply open Google Maps, enter the location, switch to satellite view, and then look for familiar landmarks nearby” (P2).

“Sometimes it shows the wrong direction, but then I look at what buildings are on the map, compare them a bit, and then I get a sense of direction and follow the route” (P1).

All participants incorporated landmarks into their design sketches, albeit with varying approaches. Some suggested displaying all landmarks along the route, while others proposed showing only the next landmark to minimise distraction. A novel “Bubble” concept was introduced, where interactive 360° street view pop-ups would appear at each turn (see Fig 2). This idea was well-received and further developed:

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Study 3: Sketches from the workshop.

(left) landmarks at important turns called “bubbles”, (middle) choosing different walking speeds, (right) a small arrow to indicate the direction towards the destination.

https://doi.org/10.1371/journal.pone.0340711.g002

“I also find it important to show Bubbles at critical locations. For me, I would also like to be able to easily drag and drop Bubbles onto the user interface” (P1).

P5 added that it would be helpful to drag the Bubble across the map to get the 360° Street view everywhere the user wants. However, P4 raised a concern:

“[A user] might miss a landmark that they wanted to see, and it restricts the navigator unnecessarily” (P4).

In the final prioritisation, the “Bubble” feature emerged as the most favoured. In total, landmark-related features received the overall highest prioritisation, underscoring their perceived importance in enhancing the RP navigation mode.

ETA Updates: Throughout the workshop, the idea of estimated time of arrivals evolved into a nuanced discussion about intention-based navigation. While participants didn’t explicitly mention ETA updates in their daily navigation habits, the feature sparked creative ideas during the sketching phase. P5 introduced a novel concept that went beyond simple ETA updates: P5 introduced a feature where users get different routes and ETAs based on 4 different walking speed options (see Fig 2). A user who chooses slow walking for example gets a route that guides them through the most scenic way, while a user who chooses speedwalking will get the fastest route. This idea was well-received, with P2 commenting:

“Usually, there is a direct route and one that might be a bit slower, similar to what you know from driving a car, but I find these walking speeds very interesting.”

P2 also suggested that alternative routes for different speed settings should be visible, allowing users to make informed decisions based on the path. Furthermore, they commented on how the map could be designed based on the users intention of discovering the environment or wanting to efficiently reach a destination:

“I have also seen landmarks primarily as orientation aids, in the sense that I only need them at a turn and not otherwise. However, there may also be the possibility of having this as an option. For example, they could be displayed by default when I need to turn, but also that I could filter by interests. If I enable discovery mode, additional landmarks may appear based on my interests. You should be able to switch between an orientation mode and a discovery mode.”

This statement shifted the focus from speed-based navigation to intention-based navigation. It combined elements of P5’s speed option idea with a new emphasis on user intentions, differentiating between goal-oriented navigation and exploratory wandering. Further refinement led to the “Discovery or Speed” feature, where users choose between the fastest route (Speed mode) or a route showcasing interesting landmarks (Discovery mode). In Discovery mode, a filter would match personal interests, displaying only relevant landmarks. Speed mode, conversely, would show only navigation-critical landmarks. While some ideas, such as using a voice assistant for ETA updates or implementing a progress bar, were proposed and subsequently discarded, the “Discovery or Speed” feature gained significant traction. It transformed the initial concept of ETA updates into a comprehensive navigation approach that adapts to user intentions, with ETA serving either as the primary focus or as background information depending on the chosen mode.

Map Orientation: The discussion on map orientation revealed a tension between cognitive load and spatial awareness. During the reflection on personal navigation habits, a clear preference emerged. Four out of five participants (P1, P2, P4 & P5) described using RP with a north-up map orientation. P5’s description exemplifies this approach:

“I open Google Maps, enter the café or place I want to go to, and search. Then I check if it is the right café that I want, and if it is, I click on the route. It usually suggests several routes, and then I see if it’s correct, and then I head in the direction of the route. I don’t click on the start button. Instead, I occasionally take out my phone, check if I walk in the right direction or see if I’ve gotten lost.”

Only P3 reported using the turn-by-turn (TBT) mode with automatic map rotation, citing difficulties with spatial orientation:

“I don’t really find my way around places, even those I know well. So, for me, it’s the same that whenever I walk from the university to the train station, I always have navigation turned on”.

The sketching phase revealed diverse approaches to map orientation. While P2, P3, P4, and P5 opted for a north-up map in bird’s-eye view, P1 proposed a 3D mode as the main navigation tool, with the option to switch to a north-up bird’s-eye view. From this concept emerged a mini-map feature. P1 described a user interface with two mini-maps at the bottom: one showing the entire route in bird’s-eye view with the user’s position indicated, and another displaying a fraction of the route. This idea was well-received, with P2 declaring it “very intuitive and thus a good addition.” P2 proposed a hybrid approach, suggesting the addition of a TBT arrow to the RP mode (see Fig 2):

“Right now, it’s a map oriented to the north, so I still have to adjust a bit. Even though I don’t really like the TBT mode, I would still like to have an arrow that shows me [...] when to turn right”.

This concept was further developed by P1 and P5, who suggested displaying the arrow on the phone’s lock screen and having an option for the arrow to always point towards the final destination. During the refinement phase, the group reached a consensus that a mode switch between north-up and 3D views might be beneficial, though the default setting remained undecided. The mini-map concept was streamlined to a single map, which could also serve as a progress indicator. In the final prioritisation, participants showed a clear preference for the mini-map concept and the arrow/compass idea. Interestingly, an initially proposed AR feature, despite initial excitement, was discarded due to a lack of interest. These results, while based on a smaller sample size, provide valuable exploratory findings suggesting a user preference for subtle, informative aids like mini-maps and directional indicators over more immersive but potentially overwhelming features like AR.