A Critical Assessment of Steenbergen’s Discourse Quality Index
The DQI consists of seven coding categories, intended to capture the principles of Habermas’ discourse ethics in terms of observable features in speech. The DQI does not attempt to measure authenticity since Steenbergen et al suggest there is no observable indication of sincerity that is not highly speculative. Steenbergen et al feel that the DQI is nevertheless a helpful measure of deliberative quality. The unit of measure for the DQI is a speech (any uninterrupted utterance). Coders recorded 56 speeches in the debate, while the coding categories relate to the content of each individual speech (Steenbergen et al 2003).
The first coding category identifies participation. It uses two codes, (0) and (1): (0) records an interruption, while (1) records normal participation (or no interruption). For an interruption to be recognized by a coder it must be explicitly acknowledged and objected to by a participant. The second coding category identifies levels of justification. It uses four codes based on the linkage between premise and conclusion – coding instructions break down as follows: (0) no premise (just a demand); (1) inferior justification (premise provided but no linkage); (2) qualified justification (linkage is made); and (3) sophisticated justification (number of linkages made between premises and conclusions). A linkage can be explicit or implicit; an implicit linkage is recorded if coders are convinced that all participants are aware of a linkage not explicitly made because it is obvious. The third category identifies appeals to the common good, with the following coding instructions: (0) one or more group interests are appealed to; (1) no group interests are appealed to; (2a) the greatest good for the greatest number is appealed to; (2b) the greatest good for the least advantaged is appealed to. These codes are not considered mutually exclusive, and could be coded in one speech. The DQI codes three further categories covering respect, and also a final category covering attempts to reach consensus. The coding instructions follow a similar pattern; respect is coded according to explicit positive or negative statements towards groups, demands, and counter arguments, while consensus is coded according to the presence of mediating proposals in speech (Steenbergen et al 2003). This paper shall not provide full details, since the coding examples given above capture the general method that this paper assesses, and with which these final examples are consistent. Steenbergen et al suggest that it is possible to derive a scale from these codes by which to measure the quality of discourse. An example of the DQI results is given concerning three of the indicators: the median, mean, and standard deviation of the coding results are provided, and can be compared to the minimum and maximum possible result of the codes (Steenbergen et al 2003). This is how discourse is measured.