5.1 P2P Online Negotiations in the Online Used Car Market
We asked (RQ1): Which behaviors do participants engage in during online P2P negotiations in the used car market? The buyers most often engaged in challenging the sellers’ competence or expertise, while the sellers often corrected and sorted out buyers’ assertions. Also, both sides appealed for sympathy or flattery while trying to establish or strengthen interpersonal bonds. Aggressive moves were omitted, which indicates that the parties were exploring the potential for a bargain without trying to enforce it. Further, it suggests that the negotiations were about matching the price, the object, and the parties’ expectations or possibilities.
Regarding the temporal structure of negotiations, we observed several sequential patterns with inherent relationships in the traditional used car market, indicating a rather high conversational complexity. We could identify three dominant patterns.
Challenge and withstand describes a buyer's behavior who, despite correcting or diverting turns from the seller, carried on with their original criticism. The data indicated that this pattern emerges in the early phases of the negotiation. We claim that the buyer tries to establish a strong negotiation position before making a concession regarding, for instance, the price.
Gain closeness describes the behavior of the parties where flattery is responded to with a similar move. This allows the parties to strengthen the mutual social connection and commitment during a progressing negotiation after clarifying the initial positions. Finally,
friendly enforce occurred in a later negotiation phase when one party uses lengthy, possibly pointless exchanges, and the other party wants to reach a clear conclusion. In this case, the other party may use BATNA to enhance the pressure on their negotiation partner. While the literature was concerned with which moves and turns occur in negotiations (Kolb
2004) our analysis unveiled typical patterns describing the sequential interdependencies between the moves and turns. The identification of typical patterns may contribute to recognizing the state of the negotiation and the conversational complexity. The conversational complexity can be visualized by the model of negotiation behaviors in the traditional used car market (Fig.
5) that shows the patterns and their relationships. We observed complex relationships between the intensively used tactical negotiation actions (high number of arrows), indicating that participants chose quasi-randomly between many different actions (Gibbs and van Orden
2012). These complex relationships make it difficult to predict what action to expect next in the negotiation process (Laubert and Geiger
2018).
Regarding deception, the results indicate little conversational honesty (cf. Gaspar et al.
2019). Although the parties did not resort to force and aggression to agree, the sellers simply lied to buyers to reach a favorable outcome—another type of immoral behavior. The interviewed participants made it clear that this was primarily a deliberate, conscious action, suggesting that people generally consider some sort of deceit to be acceptable in online price negotiations. Thus, the intuition that trusting the seller of a car is a bad idea is fairly accurate. It adds to the effect described by Akerlof (
1970) by confirming that dishonesty is an inherent part of selling a used car.
However, as illustrated by the most recent results, authenticated data may offer a way to tackle this problem. We asked
how do participants adapt their P2P online negotiation behaviors when accessing authenticated data about an offering. The major movements remained the same, confirming the overall trends identified in the conventional settings: buyers engage in challenging the sellers, who respond by correcting or diverting, and both parties try to establish social bonds by flattery. We also did not observe significant changes concerning aggressive or forceful behavior. However, the data indicated the enhanced use of diverting by buyers. And, most prominently, we could identify a new move that we refer to as the
CarCerti argument. This move indicates how easily users adopt new information provided during the negotiation into their negotiation behavior and use it to support their argumentation. The high uptake of this information confirms that information is the fundamental resource in negotiation (Lewicki et al.
2016). Being able to refer to it gives the parties new tools to engage in a more intense conversation.
The temporal and sequential analysis of the moves and turns revealed changes of the sequential patterns and their relationships and therefore a change of conversational complexity. First, we observed a specialization of the pattern
challenge and withstand toward the
severe challenge and withstand pattern. The availability of additional information provides the buyer, who otherwise would have to refer to their intuition or the overall market situation, with clear-cut arguments about a car. It is foreseeable that buyers will use a CarCerti argument in connection with the challenging move of this changed pattern. Second,
missing counterargument is a new pattern describing behavior that occurs when a buyer challenges a seller by referring to the authenticated data (i.e., a CarCerti argument), and the buyer then reacts by shifting the topic. It also explains why buyers engaged in
diverting more than in the conventional setting. It also shows the power of authenticated data as an argument that leaves the buyer no other option than to move away from the original aspect they challenged. This movement is predictable to a certain extent. Third, the patterns
gain closeness and
friendly enforce disappeared which reduces the relationships between the prominent tactical actions and therefore the conversational complexity. Thus, the revealed changes provide a better prediction of the negotiators’ behavior compared to conversations under conditions without authenticated data. Moreover, the reduction of patterns and relationships shows that the presence of authenticated data can reduce the number of options considered reasonable, and therefore conversational complexity (Gibbs and van Orden
2012). The changed patterns and reduced relationships (number of arrows) are recognizable in Fig.
7. Specifically, the observations confirm the positive relation between the authenticated data, the increased availability of correct and credible information in the conversation, and the reduction of conversational complexity as stated in our conceptual model (Fig.
1).
We can identify three essential shifts concerning conversational honesty:
change to truthful data,
change to price deception, and
change to evasive behavior. Change to truthful data reflects the reduction of information asymmetry about a car. The sellers can no longer lie about a car’s condition because it is possible that the buyer has seen or can see the authenticated data and that a lie about this would show the seller to be untrustworthy.
Change to price deception refers to sellers’ disproportionate inflation of a car’s price compared to the CarCerti data. We claim this behavior occurs because the sellers can no longer present a car more positively with false data. Hence, they choose to use the price to signify the car’s quality and hope that buyers won’t use CarCerti for comparison. Finally,
change to evasive behavior refers to the increased use among sellers of
BATNA lies. Given more information about a car, buyers may tend to engage in lengthy negotiations to reach a better outcome. Still, sellers may wish to complete a negotiation early before all potentially risky facts about a car are discussed. Overall, the identified dishonesty patterns showed that dishonest behaviors remain, but they shift in the goal—the lies were not about the car but other aspects. While this again suggests that a certain level of deceit is accepted in online negotiations, the presence of authenticated data gives more power to the buyer, who—if they wish to—may detect some dishonesty and may effectively counteract it. Thus, we can conclude that authenticated data impacts conversational honesty. Yet, the observations show a more complex picture than we assumed in our conceptual model (Fig.
1). Specifically, the participants are more honest about technical facts and figures covered by the authenticated data, but they continue to lie about other aspects, such as the status of their parallel negotiations (BATNA). Thus, rather than a reduction of dishonesty, we observe a shift between various types of deception.
The findings support previous research on peer-to-peer negotiations in various contexts. Despite the clear incentives oriented at the price of the car, many participants positioned other issues as relevant for the negotiation, like sympathy or expertise. They were discussing them as part of the negotiation. It became part of the negotiation to present and convince the other party of one’s own expertise. This created a sort of multi-issue negotiations, which was earlier shown as supporting win–win outcomes as opposite to zero-sum outcomes (Galinsky et al.
2016). For instance, one party received the acknowledgment of their expertise while the other could benefit from a lower price. Interestingly, the accessibility of authenticated data did not change such behaviors: aspects of sympathy and expertise were still coming on the table, yet the data could be used to express or confirm one’s own expertise. Following Hofstede et al. (
2019) one can confirm that a negotiation is about more than pure economic rationality. Even in an online game environment, the players try to maintain positive relations. One can interpret some cases of BATNA as an attempt to suspend unproductive negotiations without affronting the other party. This is despite the clear economic incentive structure and the risk of potentially losing the chance to reach a good deal in a parallel negotiation by spending so much time. Nevertheless, we confirm that the revelation of trusted and accurate information, even if this does not occur via the negotiator themselves as in previous research (Citera et al.
2005; Yu et al.
2021) but through a third, independent party, increases fact-related honesty and reliability of negotiation conversations. We claim that those findings hold for peer-to-peer negotiations in typical high-value transactions like in the used car market (Paese et al.
2003; Hofstede et al.
2019) or for example also in the real estate market (Galinsky et al.
2016).
5.2 Implications for Design and Behavior Adoption
This study has contributed to the negotiation literature by identifying which behaviors emerge in online, text-based P2P sales. The results suggest that participants in an online car market are driven by profit and do not refrain from deceit to reach the desired outcome. Thus, effective ways to reduce cheating are highly valued and could improve the interaction quality in online platforms—not only in used car markets but also in freelancer hiring platforms such as Upwork or online markets features on social media such as Facebook, where the participants also negotiate prices via text-based messaging. Although these platforms verify personal data or basic information concerning the property to guarantee basic trustfulness or provide comments concerning previous transactions, the de facto negotiations and deals rely on chat-based exchanges. By including authenticated data, they could enhance the value of the channels they offer and reduce the risk of deception, improving the users’ trust in a platform. Further, by decreasing conversational complexity, they can attract more users. Since many people dislike complex negotiations and consequentially tend to avoid them, introducing authenticated data to reduce negotiation complexity can provide a promising approach for platform providers to increase their attractiveness.
However, the consequences for platforms are even more far-reaching. In particular, the introduction of authenticated data can lead to people also negotiating high-value goods via online P2P platforms. The number of customers of the platforms can increase further. In addition, customers will have more choice, as they will no longer be geographically bound, for example, but will be able to buy their high-value goods without being present at the product. Up to now, they have often still been geographically bound because they want to look at and feel the good (e.g., a car) as a product to touch. Studies of these developments will become increasingly important.
The analysis revealed that, in many aspects, chat-based negotiations resemble face-to-face ones. In both, deceit and lies play a key role (Kolb
2004; Lewicki et al.
2016; Korobkin
2020). They include moves such as
challenging competence or expertise or
appealing for sympathy or flattery and rely on information asymmetry and a specific amount of mistrust. However, there were differences. For instance, the analyzed chats had almost no emotionally loaded moves such as
demeaning ideas,
criticizing style, and
making threats. A naive interpretation would be that online negotiations are less heated than face-to-face ones.
Overall, comparing the treatments indicated that manipulating the communication channel by introducing authenticated data strongly impacted the participants’ behaviors. This aligns well with the literature on communication channels’ impacts on communication characteristics (Bødker and Andersen
2005; Johnson and Cooper
2015; Kurtzberg et al.
2018). However, while the literature has focused on comparing various communication or media types, we have indicated that particular technology features change communication even if the medium remains the same.
Conventional P2P markets are strongly affected by information asymmetry (Akerlof
1970) and distrust between the parties; both aspects were also identified in the CarMarket game. A lack of information by one person about an offering offers great potential for the informed party to be dishonest and untrustworthy. Although measures against information asymmetry exist (Valley et al.
1998; Dimoka et al.
2012; Blundell et al.
2019), these are not appropriate in online negotiations. We have shown that providing authenticated data makes participants more honest about a car’s features and possibly reduces the necessity for interpersonal trust between the negotiating parties. Despite intense research, the mechanisms involved in establishing trust in these settings remain underexplored (Naquin and Paulson
2003; Söllner et al.
2012). We have proposed a way to, at least to some extent, circumvent the issue of trust toward the seller by suggesting how it can be replaced by trust toward the data and the platform. The availability of authenticated data may even enhance trust toward the seller: By being able to confirm their claims about a car with authenticated data, sellers could present themselves as more trustworthy. This increase in trustworthiness calls for further research concerning the trust mechanisms involving humans, machines, and data.