A study published in 2006 sought to examine the effectiveness of the Elaboration Likelihood Model (ELM) of persuasion to examine various artificial intelligence responses to online consumers.

The putative consumers were categorized into four groups. First, they were arranged by their “need for cognition.” Some people were determined to have a “high need for cognition”; some had a “low need.” Second, they were arranged in terms of their prior knowledge concerning the product.

The underlying premise was the two routes to decision making: central route, which is primarily cognition; and peripheral route, in which a purchaser puts more concern on emotional responses to the product or the salesperson.

These four groups of people were presented with two distinct sales interactions: One variant was the Product-Attribute-Relevant (PAR) sales pitch. This variant provided data about the product. In the experiment, they were selling cars. This would be quality, features, et cetera. A second variant  Product-Attribute-Irrelevant (PAI) strategy: this would be all of the “tricks” of the sales personnel. For example, “foot-in-the-door”:  This happens when you are presented with a small request by the seller. For example, when you call into a company and they ask you for your name and an alternative telephone number. If you have learned the persuasion categories made famous by Cialdini, then you know these non-cognitive techniques.

You have created this minor relationship status which creates some future willingness to concede further.

The results of the experiment were what one would expect.

For people with a “low need for cognition,” the increase in data did not effectively persuade them. Interpreted in light of the ELM model, that means that such people “do not have enough motivation or ability to process messages about product features.” Stated otherwise, they simply did not care. Not surprisingly, they sales-technique strategy was effective with these peripheral route processors.

Conversely, those who wanted cognition, wanted data.

But something interesting was observed: the sale-technique strategy (PAI) had about the same effectiveness on high-and-low-need-for-cognition purchasers. In line with the ELM theory, the non-cognitive techniques are easiest to process and thus everyone processed such information.

Another interesting result concerned the degree of change among various purchasers. The PAR strategy had a greater effect upon the high-need-for cognition or prior knowledge purchasers, than did the PAI strategy on the low-need-for-cognition or no-prior knowledge purchasers.

But what does one do with these results? How does a sales-agent determine the need-for-cognition of a customer quickly and without arousing suspicion that this a sales-technique?

A second experiment concerned bargaining procedure. This experiment used two bargaining techniques a tough attitude (start high, give small concession) or a moderate strategy. The purchasers had either a low or degree of prior knowledge as to what the ending price should be.

For purchasers who had little prior knowledge, the tough strategy was most effective and it did not produce a negative attitude in the purchaser. However, for the purchaser with a high degree of knowledge, the tough strategy did produce a negative attitude in the purchaser and was not more successful than the moderate strategy.


Huang Shiu-li, Lin Fu-ren, and Yuan Yufei, “Understanding Agent-Based On-Line Persuasion and Bargaining Strategies: An Empirical Study,” International Journal of Electronic Commerce 11, no. 1 (Fall 2006