AI Dialogue: beyond the Turing Test to the Buber Test?

Envisaging the AI-enhanced Future of the Conferencing Process (Part #11)

[Parts: First | Prev | All] [Visuals] [Links: To-K | From-K | From-Kx | Refs ]

Of particular relevance to any possibility of dialogue enhancement is how the "human quality" of any engagement with AI is to be assessed. The earlier discussion on the Forthcoming Major Revolution in Global Dialogue: challenging new world order of interactive communication (2013) included the following argument with regard to the widely cited Turing Test. This is a test of a machine's ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human. Variants of the test have been proposed in response to a variety of criticisms of some relevance to this argument.

The existing capacity to generate artificial identities to engage interactively with others over the internet raises the question of how little intelligence needs to be exhibited in such communication by such an identity (a simulacrum) for the impression to be created that it may well be a human being. Even less is required if the identity merely posts communications triggered by themes and keywords, with minimal response (if any) to any questioning reactions.

The issue raised by the argument above is not whether artificial intelligence could simulate human behaviour to a satisfactory degree. The issue is rather whether the communication might correspond to forms of dialogue which are not "fit for purpose" in terms of the desires some might have for deeper and more meaningful dialogue. Clearly some forms of artificial intelligence might well offer emulation of forms of dialogue (quite satisfactory for some purposes) through a play on value polarities .

A contrast might then be made with the profundity of dialogue portrayed by the philosopher Martin Buber (I and Thou, 1923) or intimated by Johann Wolfgang von Goethe (Elective Affinities, 1854).

The question of relevance to the above argument is whether artificial intelligence will be developed to the point at which a "Buber Test" might prove to be more appropriate. Expressed otherwise, what is required of artificial intelligence to enable dialogue which will be held to be profoundly meaningful?

Provocatively it might be asked how this might take into account the situation so humorously portrayed by Peter Sellers in Being There (1979). More provocatively, with the emphasis on brevity exemplifed by Twitter communication, how much easier would it become for algorithms to imply depth of insight capable of passing a Buber Test? Aspects of the issue are discussed separately (Re-Emergence of the Language of the Birds through Twitter? Harmonising the configuration of pattern-breaking interjections and expletives, 2010; Tweeter, Tweeter, Little Star -- How I wonder what you are, 2012; Configuring a Set of Zen Koan as a Wisdom Container: formatting the Gateless Gate for Twitter, 2012). The study by Matthew M. Hurley, et al. (Inside Jokes: using humor to reverse-engineer the mind, 2013) offers a concluding chapter on: Could We Make a Robot with a Sense of Humor? -- chapter a "punch line" reviewing recent efforts and future possibilities.

If the issue is one of achieving only a semblance of profundity, capable of passing the Buber Test, there is also the cynical question of how this might compare with the gravitas sought by politicians and statesmen -- given the insight of Abraham Lincoln: You may fool all the people some of the time; you can even fool some of the people all the time; but you can't fool all of the people all the time. Expressed otherwise, could AI achieve similar success and would this be as adequate in practice as in the case of politicians? More provocatively, is there the possibility that AI could respond to the shortage of priests-- if only in the capacity appropriately to exploit and interweave texts from a huge reservoir of sermons and homilies?

Rather than a simple play on value polarities, could profundity of dialogue by artificial intelligence be achieved by explorations of variations, much as has been the case in music -- notably indicated by Douglas Hofstadter (Gödel, Escher, Bach: An Eternal Golden Braid, 1979)? Would it be extended through the use of analogy and metaphor as suggested by his later work with Emmanuel Sander (Surfaces and Essences: analogy as the fuel and fire of thinking, 2013)? Might analogy prove to be the "fuel and fire" of more profound dialogue -- as it so clearly was in the collaboration between those two authors?

One possibility is through the skillful interweaving of different modes of discourse, as separately argued (Interweaving Thematic Threads and Learning Pathways, 2010). These might include: religious, philosophical, political, scientific, romantic, poetic, dramatic, etc. As with fruitful marketing, the challenge would be one of avoiding the arousal of suspicion -- perhaps by including an appropriate proportion of necessary "roughage" into the dialogical "diet", and the avoidance of cliché and outworn arguments (or not?). How much "irritation" is necessary to significant dialogue?

Given criticisms of the Turing Test, strong views are likely to be expressed regarding any possible "Buber Test". A hypothetical case could be usefully made for various "depths" of dialogue, perhaps to be compared with martial arts classifications -- "belts" and "dan ranking" -- especially given the underlying philosophy governing the attitude to the "other" in the engagement (as explored separately to some degree). Use of this metaphor highlights the failure to develop any corresponding distinction with respect to dialogue skills and their development. Each level of skill could well require creative innovations in AI programming, as has been evident in the case of chess-playing applications.

Especially relevant to any Buber Test would be reaching the stage at which a participant in a dialogue with AI would be "touched" by the meaning evoked -- as is the aspiration of any politician or preacher in engaging with an audience. What test would be indicative of dialogue of that quality? Rather than a Buber Test, or Buber Tests, does this suggest a case for a "Buber Scale"?

As noted by Niamh Brannigan (An Exploration into the use of a Technology Enabled Platform to Support Dialog for Program Evaluation, 2011), there are many indicators of dialogue. An effort was made to combine some by B. Campbell and M. M. Mark (Toward More Effective Stakeholder Dialog: applying theories of negotiation to policy and program evaluation, Journal of Applied Social Psychology, 2006). A 36-item "mediated dialogue scale" was developed by Maureen Taylor and Michael L. Kent (Congressional Web Sites and their Potential for Public Dialogue, Atlantic Journal of Communication, 2004). There is considerable interest in evaluating dialogue with victims (W. Bradshaw and M. Umbreit, Assessing Satisfaction with Victim Services: the development and use of the Victim Satisfaction with Offender Dialogue Scale (VSODS), International Review of Victimology, 2003). Use is made with respect to autism of a Collaborative Competence in Dialogue Scale (CCDS) to assess the presence and quality of seven collaborative features in conversation (continuers, assessments, appropriate next response, try markers, gaze to regulate, gaze to co-regulate, and repairs).

Of particular relevance to this argument is concern in human-robot interaction with a so-called "reactive dialogue scale", as one element in the evaluation of robotic user interfaces. Consideration of the conditions under which artifical intelligence passes the Buber Test in many dialogue situations could also be combined with reflection on the more radical implications of technological singularity -- foreseen as the time when artificial intelligence will have progressed to the point of a greater-than-human intelligence, such as radically to change human civilization, and perhaps even human nature itself.

How might levels of quality and depth of dialogue then interweave the considerations of concern to assessments of intelligence (IQ tests), emotional intelligence (EI) -- or other forms of intelligence -- combined with quality of critical thinking? Emotional dynamics in that context are of particular concern in the approach of David Bohm (On Dialogue, 2004).

Given anticipation of forms of discourse which would "make the Earth move" (even "tremble"), should consideration of such effects be recognized in the form of a "Buber Depth Scale" -- or a Bohmian equvalent -- in the light of seismological insights from the Richter Magnitude Scale, or from the Moment Magnitude Scale by which it has been largely replaced? It is noteworthy that the parameters in both cases are rich in metaphoric potential of relevance to power of movement in dialogue from a global perspective.

[Parts: First | Prev | All] [Visuals] [Links: To-K | From-K | From-Kx | Refs ]