Direct vs. Score-based Selection: Understanding the Heisenberg Effect in Target Acquisition Across Input Modalities in Virtual Reality

TL;DR

This paper analyzes the Heisenberg Effect in VR target acquisition across input modalities (controller vs bare-hand) and selection mechanisms (direct vs score-based) via a within-subject study. It finds hand input is more susceptible, with width driving the effect for direct selection and target density driving it for score-based selection; a temporal history model, Weighted VOTE, is proposed to mitigate disturbances by weighting prior intention samples, significantly improving selection accuracy especially for bare-hand inputs. The work provides a principled, modality-spanning view of the Heisenberg Effect and demonstrates a practical mitigation strategy that can adapt to individual user behavior. The findings have implications for designing robust VR interaction techniques and motivate future work on adaptive, history-informed selection across diverse hardware and contexts.

Abstract

Target selection is a fundamental interaction in virtual reality (VR). But the act of confirming a selection, such as a button press or pinch, can disturb the tracked pose and shift the intended target, which is referred to as the Heisenberg Effect. Prior research has mainly investigated controller input. However, it remains unclear how the effect manifests in the bare-hand input and how score-based techniques may mitigate the effect in different spatial variations. To fill the gap, we conduct a within-subject study to examine the Heisenberg Effect across two input modalities (i.e., controller and hand) and two selection mechanisms (i.e., direct and score-based). Our results show that hand input is more susceptible to the Heisenberg Effect, with direct selection more influenced by target width and score-based selection more sensitive to target density. Based on previous vote-oriented technique and our temporal analysis, we introduce weighted VOTE, a history-based intention accuracy model for target voting, that reweights recent interaction intent to counteract input disturbances. Our evaluation shows the method improves selection accuracy compared to baseline techniques. Finally, we discuss future directions for adaptive selection methods.

Figures (5)

• Figure 1: An example from our experimental design in Unity, which includes nine combinations of target width and spacing. In this example, the target width is 42 cm, and the target spacing is 70 cm. The distance between the user and targets is fixed at 8 meters.
• Figure 2: The performance summary of each technology under different target width conditions, including selection time, overall error, Heisenberg Error and Heisenberg Magnitude. The error bar represents the standard error. DH: direct selection with hand input, SH: score-based selection with hand input, DC: direct selection with controller input, SC: score-based selection with controller input.
• Figure 3: The performance summary of each technology under different target spacing conditions, including selection time, overall error, Heisenberg Error and Heisenberg Magnitude. The error bar represents the standard error. DH: direct selection with hand input, SH: score-based selection with hand input, DC: direct selection with controller input, SC: score-based selection with controller input.
• Figure 4: Distribution of selection endpoints across input techniques with a target width of 28 cm and spacing of 70 cm. Endpoints under direct selection (a, c) are more concentrated around the targets, whereas score-based selection (b, d) shows greater dispersion. These endpoints represent the raw pointing ray rather than the final snap-to raycasting.
• Figure 5: Trends of user intention accuracy over time across four selection modalities and mechanisms. Each curve is fitted with a third-degree polynomial within 0.4s before selection. Relative time (x axis) is normalized across selection events, where 0.0 denotes selection completion. Accuracy of Intention (y axis) is defined as the probability that the user's selection ray is pointing at or snapping to the correct target object, where 1.0 denotes that the intended object must be the target one. The first row shows overall accuracy curves trained on data from P1–P19, illustrating general trends. The bottom row highlights individual participants whose accuracy patterns deviate from the overall trend.