On the Feasibility of Remotely Administered and Self- Driven Cognitive Assessments using a Multimodal Dialog Agent