Affordance and Symmetry in User Interfaces

to have any psychological validity. Further work is called for, and we suspect a fertile area will be in defining homomorphisms between the groups of different system representations: for example, between the physical user interface and the structure of the transition system ‘underneath’ it. There is also a mischief element in human nature that our examples can highlight as another form of affordance. Having had a particular action made salient by a symmetryaffordance, the mischief question then becomes, when can I abuse the symmetry-afforded action to break symmetry? For example, in the tangible interface example, the edge of the table is clearly distinct from the rest of the table, so what happens if you put a cube there? Would not it be funky to have a clock that does not sweep out equal angles in the same time periods? Would it be possible to have staircases that do not afford climbing? Indeed, this last question is the basis of tricks and deliberately designed challenges in games such as the Tomb Raider series where scenery is intended to be used for climbing even if it does not at first glance appear that way. These questions go beyond the two principles of symmetry-affordance defined here into how other actions become salient as a consequence of symmetry and a mischievous nature. They may equally well be affordances in the more general sense but they would not be symmetry-affordances. 8. SUMMARY AND RECOMMENDATIONS FOR APPLYING SYMMETRY To summarize, we claim that part of the psychological affordance inferential mechanisms (which we called symmetryaffordance) is that symmetries suggest actions that a user can do, and moreover that those actions respect the same symmetries. A designer can exploit this to ensure that users’ expectations are met. To apply the notion of symmetric-affordance in design, it is important to consider the key elements of the model. There are three key components: (i) the symmetries of the system, that is, transformations that leave the system looking the same; (ii) symmetries in the perception of the system; and (iii) the inferential mechanism that constitutes affordance. The first point is the standard Weyl definition of symmetry. The second point arises because affordance requires consideration of the relation between physical objects and mental representations. We are not aware of the important role of the model being raised in any other discussion of symmetry. As has been seen though, an explicit inferential mechanism, even as lightly specified as the one here, is enough to provide insights into affordances. It is possible to consider deeper models that may base their inferences on things other than symmetries. As we have seen, learning and previous experiences can help to formulate the inferential model but exploiting these in a way as immediate as symmetry is not apparent to us. We now summarize insights we have made in passing throughout this paper. (For brevity we do not revisit the mischievous and challenging potential—even ‘conventional’ uses such as exploiting the security and educational potential of deliberate obscurity—of some design decisions, covered in Section 7.7.) 8.1. Insight 1: symmetries indicate actions Symmetric parts of an interface, particularly a physical interface, suggest actions because of the immediacy of the symmetry and the transformation associated with the symmetry. Thus a symmetry in the interface can indicate to the user’s inference what actions are possible. A designer can exploit symmetry by ensuring that symmetric transformations really do correspond to meaningful actions on the system. A classic example of this is the ‘wheel’ on an Apple iPod. Wheels have rotational symmetry; this suggests that the user could try turning the wheel and, fitting with this, turning the wheel does indeed help to control the interface. 8.2. Insight 2: broken symmetries indicate necessary feedback Some transformations of the system need not be symmetric, for instance, a battery has died. In these cases, if the user does not perceive the difference caused by the transformation, they are going to infer actions that are potentially invalid, given the new state of the system. At first glance, this may seem to lead to the well-known and well-worn usability guideline of providing the user with feedback [27]. However, with symmetry we are able to say more: if the system changes but the inferred set of actions is not different, then feedback on the system state may not be necessary. Thus, we can use symmetry as a way of deciding those transformations requiring attention (and so employ pop-out, colour changes and the like in the user interface) and those transformations that do not (and can simply result in a change of display that can safely be ignored). 8.3. Insight 3: symmetric actions should be permissive Symmetrically prompted actions also have natural constraints. As discussed above, because of the consistency of inference of actions, symmetric actions should be permissive (commute with all other actions). Again, this need not be the case but in fact meeting with this expectation would support the symmetric affordance of the interface. Additionally, a symmetrically prompted action should always be effective as it is always cued for by the interface. Page 10 of 12 P. CAIRNS AND H. THIMBLEBY THE COMPUTER JOURNAL, 2007 Interestingly, on the iPod, turning the wheel clockwise whilst at the bottom of a menu does not change the interface. Here, the symmetry-affordance has been broken and the resulting action is ineffective. There may be good reasons for this to do with scrolling through long lists, but symmetry-affordance also suggests that the designers might not have got this right. 8.4. Insight 4: symmetries account for modelessness and consistency When systems look the same regardless of time or location, they should behave the same. These attributes are already understood as the concepts of modelessness and consistency and thus are in fact specific cases of symmetry-affordance. They do not need dealing with separately but can be analysed, in the same way as other symmetry-affordances, by consideration of the underlying temporal and spatial translations. 8.5. Insight 5: the need for validation Formal models stand or fall on their validity, which is a mathematical question. A proposed model of affordance further needs to be valid psychologically, and in some sense improve the quality of design or the quality of the design process. Unfortunately, validating symmetry-affordance in this sense strictly requires an appropriate experimental method, and access to designers with the necessary skills to generate testable designs based on the proposed concepts. This form of empirical validation is very much further work. In the meantime, our concept of symmetry-affordance is a research contribution that has value because it helps clarify an important concept, and raises many new questions. Whether such clarification leads to better designs is another matter; indeed, if it fails to lead to better designs, it is still possible that the formalisation was valid but the very concept of affordance is not as useful for HCI as we had hoped.