Transepistemic Abduction

a specialised mode of reasoning

Transepistemic Abduction (TeA) is a specialised mode of reasoning which supports the use of intuition and heuristics when working between divergent epistemic domains.

Reflective Writing Analytics (RWA) necessitates working across two different epistemic domains, the psychosocial and the computational. This transepistemic work is problematic due to divergence in the explanations each domain affords of features in the reflective writing. The common approaches to addressing this divergence have difficulties in justifying the various decisions and assumptions involved. I address this issue by presenting a mode of reasoning which I call Transepistemic Abduction (TeA).

The objective of TeA is to provide a way for well-reasoned intuitions to be adopted for reconciling explanations from different epistemic domains. For RWA the phenomena of concern is the reflective text. The reflective text is an output from the psychosocial activity of writing, and can be explained psychosocially. This same artefact is also the input for the computational activity of symbolising and can be explained computationally. When these explanations are reconcilable and provide the required meaning and accuracy, there is no need for TeA. However, explanations frequently cannot be reconciled without requiring change to one or the other epistemic domains. It is in this situation that TeA is required.

The inability to reconcile explanations between the psychosocial and the computational stems from the limitations of the epistemic resources of those domains. Because of this, reconciliation is only possible through the use of epistemic resources outside those domains. However, this presents a contradiction, as explanations that lie outside the epistemic domains can not be justified based on the epistemic resources of those domains. That is, an explanation that does not fall within the domain of psychosocial knowledge can not be justified on the basis psychosocial knowledge. Similarly, an explanation that is based on knowledge that is not computational can not be justified on the basis of computational knowledge. What is required is a way of providing a reasonable approach to reconciling explanations despite limited epistemic resources.

For RWA, abduction offers a reasoned basis for taking action when the epistemic resources to hand are inadequate. The RWA conceptual model (see RWA) is based on two different epistemic domains. The differences between these domains result in differing explanations of reflective writing (the phenomena), and as deeper psychosocial meaning and greater computational accuracy is sought, the explanations tend to diverge and become irreconcilable. Like cognitive irritant in abduction, the irreconcilable explanations present as an irritant when working across two different epistemic domains. As with abduction, the epistemic resources available are not sufficient to provide reconciled explanations, and therefore leaving the irritant unresolved.

When abduction is applied to the two epistemic domains of RWA, action can ensue on the basis of reconciled explanations. Trans-epistemic action is possible where both domains can move forward despite their epistemic limitations. The ignorance preserving character of abduction means that this progress can be made on a reasoned basis, but without requiring change in either domain. Abduction provides the benefits of action without the cost of knowledge” (Woods:2012).

The figure above shows how abduction can be used in a transepistemic way - as Transepistemic Abudction (TeA). With abduction the hypothesis can provide the means to hit the cognitive target. In TeA however, each domain provides an explanation of the phenomena, and so the hypothesis needs to provide a relation that allows these explanations to be reconciled. It is this relation that allows transepistemic progress to occur.

A Description of TeA

The observation of a Phenomenon requires explanation in the two different Epistemic Domains ($A$ and $B$). Importantly, this is a single phenomenon that is subjected to examination utilising the resources of two different epistemic domains. It is not two variants of the one thing. For example, in RWA the phenomenon is the reflective text, the artefact that holds both a psychosocial interpretation, and computational representation and modelling.

It is the Epistemic Resources of the domains that result in specific explanations of the phenomenon. These resources can be understood as the knowledge of each domain (denoted by $K_A$ and $K_B$). Explanations are domain specific and so an explanation from domain A ($E_A$) is a consequence of the knowledge of domain A, and similarly for domain B. Therefore the explanation in domain A is an explanatory consequence of the knowledge of domain A (This can be formalised as $K_A \looparrowright E_A$). Explanatory consequence should be construed broadly since different domains have different norms in regards to how acceptable or valid explanations relate to their knowledge. Because the respective knowledge bases are different (including the means for how explanations are derived from knowledge), the domains will likely result in two very different explanations ($K_A \looparrowright E_A$ and $K_B \looparrowright E_B$) for the same phenomenon. In such cases, there are two possibilities; the explanations are reconcilable (denoted $E_a \sim E_b$), or they are not ($E_a \nsim E_b$).

Irreconcilable explanations present as a Cognitive Irritant as they cannot be resolved utilising the epistemic resources of one or both of the domains, and as a result no action can be taken. This can be seen in the previous RWA example with identity and pronoun frequency. The psychosocial resources on identity do not address parts of speech, and the computational resources on parts of speech do not speak to identity. With no justified way of reconciling these different explanations of the reflective writing, a cognitive irritant results with no clear path to action.

An Ignorance Problem exists for both domains as there is a gap between the available knowledge and the need to reconcile the explanations. Ignorance problems exist when the explanations can’t be reconciled without adding to or modifying the epistemic resources of either or both domains (i.e. subduance and surrender). For the RWA example, the explanations might be resolvable if Identity Theory included information about specific language use. Similarly, it could be resolvable if in Computational Linguistics parts of speech held specific meaning with respect to identity. However, for this to occur the knowledge bases of Identity Theory and Computational Linguistics would need to be extended with this information. Without such an extension, each domain retains an ignorance problem with respect to the other’s explanation.

The target is to reconcile the explanations (i.e. move from $E_A \nsim E_B$ to $E_A \sim E_B$) without the requirement that the knowledge bases of either domain be changed. The Hypothesis is based on a Relation (denoted $R$), which connects the respective explanations $E_A and E_B$ so as to reconcile them ($E_A R E_B$). Much hinges on the discovery of this relation. An example RWA hypothesis may involve a relation that connects the psychosocial concept of professional identity formation and computational models of first person pronouns and topics associated with professional language. This may be evidenced by changes over time; for example, if changes in the language over time mirrored an identity change over time as assessed by applying a psychosocial perspective.

If reconciliation occurs, TeA results in Transepistemic Action. Transepistemic action is possible when the explanations are reconciled, providing a way forward that is appropriate to both domains. ‘Way forward’ is an important idea as there is no prescribed goal that is known before the hypothesis is established.

As a specialised mode of reasoning, TeA is a cognitive process. When a phenomenon $P$ is observed in the context of epistemic resources $K$ (knowledge) in two domains $A$ and $B$, two different explanations $E_A$ and $E_B$ result. TeA describes the reasoning that is required in order to reconcile these explanations, thereby allowing actions such as the use of intuitions and heuristics.

This excerpt on TeA has been adapted from my PhD thesis. You can find a copy of the thesis on my QUT ePrints Page.