| Summary: | Computer agent technology is a useful technology for a customer to deal with
information overload and search complexity in ecommerce. This study focus on how a
product-brokering recommendation agent (RA) can be designed better so that a customer will
be more likely to trust and then adopt the RA for the customer's purchase decision making.
Our research model predicts that internalization and familiarity will increase both
cognitive trust and emotional trust in an RA, which will then increase a customer's intention
to adopt the RA either as a delegated agent or as a decision aid. Internalization refers to a
customer's perception of how well an RA represents the customer's real needs. A customer's
familiarity with an RA is expected to increase from the initial interaction through repeated
interactions. We conducted a lab experiment to test the model. We also used a processtracing
method to track the processes of trust formation.
The results of the quantitative analysis (PLS analysis) support our hypotheses.
(1) Internalization significantly increases a customer's intention to adopt an RA via
increased cognitive trust and emotional trust. This implies that an RA should make sure a
customer's specifications of product attributes (i.e. the expressed needs) actually represent
the customer's real needs instead just taking a customer's specifications as they are, and that
the focus of RA design should shift from product filtering toward customer representation.
(2) This study shows that RA adoption is not purely rational or cognitive. RA
adoption is affected by both cognitive trust and emotional trust. Emotional trust in an RA
directly increases a customer's intention to adopt an RA, and emotional trust fully mediates
the impact of cognitive trust on the intention to adopt an RA. The effect of emotional trust on
the intention to adopt an RA as a delegated agent is higher than its impact on the intention to
adopt an RA as a decision aid, while the effect of cognitive trust on the intention to adopt an
RA as a delegated agent is lower than its effect on the intention to adopt an RA as a decision
aid. This implies that emotional trust should be investigated in IT adoption and trust research.
(3) The intention to adopt an RA as a delegated agent and the intention to adopt an
RA as a decision aid are different levels of RA adoption that merit separate research. They
are different at the conceptual, measurement, and behavioral levels.
(4) The total effects of familiarity on the intention to adopt an RA are low, indicating
that a customer decides on RA adoption mainly during her initial interaction with an RA.
In addition, we used a process-tracing method to track how customer trust/distrust in
an RA is formed. To our best knowledge, this is the first empirical study to trace the
processes of trust/distrust formation. We developed two coding schemes in terms of: a) the
source and the destination of a process, and b) trust/distrust formation via different ways that
customers process their knowledge of an RA. The verbal protocols from 49 subjects,
speaking 40 minutes per person on average, were collected. The results of the protocol
analysis include:
(1) Trust and distrust have different structures in terms of cognitive and emotional
components.
(2) Trust and distrust in an RA are formed via different ways that customers process their
knowledge about an RA
(3) An RA needs to be designed in different ways in order to facilitate cognitive and
emotional trust while preventing cognitive and emotional distrust.
(4) Repeated interactions change neither the structures of trust/distrust nor the proportion
of different trust/distrust formation processes. However, the quantities of processes in
all categories drop by almost 50% from initial interactions to second interactions, and
then reach a stable level.
(5) The level of RA internalization change neither trust/distrust structures, nor the way
that customers process their knowledge in order to form trust or distrust.
(6) Finally, an RA needs to be designed in different ways in order to facilitate cognitive
trust vs. emotional trust. Similarly, the RA needs to be designed in different ways in
order to prevent cognitive distrust vs. emotional distrust.
|
|---|
