From watsonb@cs.ualberta.ca Wed May 26 18:59:01 1999 Received: from burdell.cc.gatech.edu (root@burdell.cc.gatech.edu [130.207.3.207]) by lennon.cc.gatech.edu (8.9.1/8.9.1) with ESMTP id SAA05704 for ; Wed, 26 May 1999 18:59:01 -0400 (EDT) From: watsonb@cs.ualberta.ca Received: from asbestos.hitl.washington.edu (hitl-new.hitl.washington.edu [128.95.73.60]) by burdell.cc.gatech.edu (8.9.1/8.9.1) with ESMTP id SAA29588; Wed, 26 May 1999 18:58:58 -0400 (EDT) Received: from scapa.cs.ualberta.ca (root@scapa.cs.ualberta.ca [129.128.4.44]) by asbestos.hitl.washington.edu (8.9.3/8.9.3) with ESMTP id PAA22030 for <3d-ui@hitl.washington.edu>; Wed, 26 May 1999 15:57:12 -0700 (PDT) Received: from localhost (user: 'watsonb' uid#105 fake: innisfree.cs.ualberta.ca) by scapa.cs.ualberta.ca id <13336-3057>; Wed, 26 May 1999 16:56:47 -0600 Subject: Re: Virtual vs. Real Manipulation To: 3d-ui@hitl.washington.edu (g 3D user interfaces) Date: Wed, 26 May 1999 16:56:46 -0600 (MDT) In-Reply-To: <199905261927.PAA24218@lennon.cc.gatech.edu> from "Doug Bowman" at May 26, 99 03:27:06 pm X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Message-Id: <19990526225655Z13336-3057+323@scapa.cs.ualberta.ca> Status: RO Hi all, So since I'm supposed to be an expert on the effects of delay etc. on human performance, I thought it best I pipe up here. You are talking here about two sorts of tasks: tracking and placement. In tracking tasks, users must move a cursor along a specified path, within certain error constraints. This task has a long history in delay research. One particularly onerous variation of this task has users "following" an unpredictable target, rather than a visual path. Typically the task has only position constraints (as opposed to orientation constraints -- hey this is hard enough). Tracking tasks translate loosely to tasks like piloting/navigation. Related research comes from people like Kenyon, Eggleston and Ellis. Steve Bryson has done some interesting pilot studies. Placement tasks are probably more familiar. In 2D and 1D there is a long history of Fitts' Law studies. Attempts to generalize Fitts' Law to 3D placement haven't been going extremely well (see in particular the work from Ware & Balakrishnan). Most of these studies also ignore orientation, though many real world tasks do have orientation constraints (e.g. a plug in the wall). One interesting line of research of late, appearing again at the recent CHI conference, is the work on time prediction for trajectory tasks in 2D by Shumin Zhai. Basically he has generalized Fitts to produce a model that can predict 2D tracking times (not error). It would be interesting to see if this could be generalized to 3D... but he's not yet working on that problem. So, end of the lecture and on with the comment -- experimental results suggest exactly the opposite of the relationship of delay to difficulty that Doug proposes. For 3D object placement (location only), some recent work I did, reported at CHI, shows that more difficult tasks are *more* affected by delay. Indeed, motion is slower, but that is because of the large number of fine adjustments that must be made. Users must make an adjustment, evaluate the results, make an adjustment... and with every such feedback loop iteration, they must also wait for delay. More difficult tasks require more feedback iterations, and therefore are more affected by delay. The inverse is true of "easy" tasks. My work does not speak to tracking, but the wealth of work by other researchers points to a similar relationship between difficulty and delay. Indeed, since tracking involves a "continous" sort of placement over time (see Zhai above), and often an element of prediction, this relationship seems to be even stronger. So, to finally burrow myself out of this hole I've dug, one would have to say that simpler tasks will be done better in VR than more difficult tasks, because difficult tasks require prompter and higher quality feedback -- and that is where reality will most often win. This might not apply if we allow "magic" interfaces -- but what they essentially do is change the task or improve the feedback. Ben. > > > 2. Path tracing could similarly be done using a "wire-and-loop" type of > > electric system. (I remember reading a paper on a virtual system that did > > this...anyone have the reference?) > > Don't know if this is the paper you're thinking of, but the > paper titled "Effects of Network Characteristics on Human > Performance in a Collaborative Virtual Environment" by Park > & Kenyon at VR '99 this year used a task like this, except > that it was done by two distributed users. > > Robert's lists of tasks bring up an interesting question. The > "Operation" task and the wire-and-loop task are both very difficult > even in the real world. Will VEs be closer to real-world performance > on these difficult motor tasks, or on very simple motor tasks? > > I think both sides of this could be argued. On the one hand, easy > tasks require little accuracy and are very forgiving, thus lessening > the effects of tracker jitter, bad depth perception, etc. On the > other hand, people generally do the very difficult tasks extremely > slowly, which could lessen the effects of lag. Thoughts? > > --Doug > > P.S. Robert, my defense is tomorrow - thanks for asking - I'll > let you know how it went. > -- > Doug Bowman, Ph.D. Candidate > College of Computing, GVU Center, Georgia Tech > Room 388 CRB, (404) 894-5104 > bowman@cc.gatech.edu > http://www.cc.gatech.edu/~bowman/ > -- Benjamin Watson, Ph.D. Assistant Professor Department of Computing Science 615 General Services Building Edmonton, Alberta, Canada T6G 2H1 tel: +1 780 492 9918 fax: +1 780 492 1071 email: benjamin.watson@ualberta.ca URL: http://www.cs.ualberta.ca/~watsonb