Head-Mounted Augmented Reality (AR) systems permit the real-time embedding of visual and textual information in the field of view of cooperating users. We start with two head-mounted AR-systems equipped with an active memory infrastructure (AMI), capable of automatically structuring multimodal memory in an intelligent way. This allows us to investigate alignment along the following lines:

1. We search for evidence of alignment via data analysis of memory traces from AR-based-cooperation experiments. The alignment recorder, a novel monitoring tool for observing the dynamics of alignment processes, was implemented to improve active memory structures in accordance with the alignment model of Pickering & Garrod (2004).
2. AR-based communication permits to actively trigger misalignment, for instance by augmenting tangible objects with contradictory visualizations for the users in cooperative sorting games. This (in conjunction with the simultaneous memory recording) allows us to induce misalignment and to study alignment processes in an otherwise impossible way. 
3. AMI-supported AR-systems offer to develop supportive alignment channels by making their memory traces mutually transparent by visual or auditory augmentations. We investigate the effects of such shared information on cooperation and their utility to accelerate or modify alignment processes.

Based on the above points, in our current approach, we investigate alignment processes in a semi-experimental task-oriented dyadic collaboration setup using AR as a "interception and manipulating device" that allows to record and precisely influence the participants’ multimodal perceptions. With respect to Pickering & Garrod our overall goal consists in getting a deeper insight in the occurrence of alignment in dialogue and in a narrower sense to gain insight in the methods interlocutors use for sharing the same representations at some linguistic level.

In the experiment, pairs of participants are seated across a table, facing each other. They are equipped with the multimodal AR-based interface and asked to jointly envision a museum exhibition using a set of objects (5cm cubes as material "handles" for augmented objects sitting on top of the cubes) on a given floor plan. The virtual objects were chosen in terms of their object-specific properties, so that some object constellations exclude each other. The aim is to enable a process of negotiation with a potential for discussion which is based on decision making in form of "Object A conflicts with B" or "Object A is in accord with B". 

In such a scenario a particular challenge for the participants consists in establishing, maintaining and manipulating joint attention both to relevant objects and to the interaction situation itself. In the study of the interplay of joint attention and alignment we see a promising avenue to investigate human multimodal interaction. In doing so, our analytic method draws upon Ethnomethodological Conversation Analysis in order to answer the following questions:

1. Which are the methods interlocutors use to establish joint attention in these specific circumstances?
2. Which communication resources do participants use?
3. Which limitations arise in the AR-setup and how participants deal with them?

Subsequently we apply data mining methods to verify hypotheses and handle a larger interaction corpus. In result this leads to the identification of features to measure alignment processes in dyadic cooperative interaction.