Digital leadership: Motivating online workers

Contingent rewards and transactional leadership

Leaders regularly use their power to reward and punish followers. For instance, they provide contingent rewards to reinforce high performance. A leader's relationship with their followers is based on a proper exchange of resources. This exchange is often referred to as transactional leadership behavior [8]. Transactional leaders offer bonus payments for good results and punish inferior performance. The provision of contingent rewards and punishment relies on efficient incentive systems and compensation plans in organizations. Contingent rewards work best in simple environments where input or output is easily measurable [7]. However, if, for instance, one aspect such as quality is not easily measurable, the provision of contingent rewards might not be the best option [9]. In most organizations, incentive systems enable leaders to reward good performance. Those incentive systems range from simple piece rates to complex bonus payments. In remote settings, these incentive systems can be used to mitigate the problem of lower social control and potential exploitation. Most platforms already have systems in place in the gig economy where leaders can grant bonus payments for excellent work or refuse to pay for poor results.

One study from 2018 considers the impact of different incentives, including contingent performance pay in a large experiment on Mturk, one of the largest online labor markets [10]. On Mturk, gig workers typically work on simple tasks, such as labeling pictures, transcribing texts, or completing surveys for brief time periods. In the 2018 study, workers had to press the buttons “a” and “b” alternately. Each pair of a-b presses generated one point for the worker. Thus, it was clear to the workers that they participated in an academic experiment and the task had no inherent meaning. The results show that using contingent rewards by offering an output-related piece rate scheme worked. Paying a piece rate led to increased performance. Whereas introducing a low piece rate of 1 cent for 100 points had a strong effect, increasing the piece rate further only led to moderate improvements. The findings align with the results from a 2021 study [3], whose authors also examined worker behavior on Mturk. In contrast to the earlier study [10], they used a transcription task with two dimensions: quality and quantity of the transcribed text. In this setup, introducing a piece rate based purely on output might result in lower quality. In both studies, providing a low piece rate increased output. However, offering a piece rate that was five times higher led to no significant increase of output compared to the low piece rate in the more recent study [3]. Interestingly, the increased output from the low piece rate did not come at the cost of reduced quality. Even if the researchers announced that they would not control for quality and thus the workers did not have to fear that their work would be rejected and not paid, the quality remained stable.

Summing up, techniques based on contingent rewards in the form of performance pay are effective in the gig economy. However, the effect of increasing payments on output is weak and leads to increased costs for the employer. In addition, it is important to keep in mind that paying output-based performance pay is not feasible if the desired output cannot be measured easily [7]. Thus, in more complex remote work environments, contingent rewards might not be the most efficient type of leadership.

Charismatic and transformational leadership

Besides contingent reward and punishment mechanisms, leaders rely on charismatic leadership techniques to motivate their followers and establish trust. Charismatic leadership is defined as “values-based, symbolic, emotion-laden leader signaling” [5], p. 304. The concept builds on so-called verbal and non-verbal leadership tactics or signals that aim to arouse emotions, thereby making the leader's message more salient. Charismatic leaders use, for instance, rhetorical questions or tell anecdotes to provide a frame and create a vision. They announce goals and express moral conviction to emphasize the substance of their message. Non-verbal techniques such as body gestures and facial expressions are often used to signal the leader's emotional state. Note that the charismatic leadership concept is often perceived as part of, or related to, the popular transformational leadership style, which encompasses: idealized influence, inspirational motivation, individualized consideration, and intellectual stimulation [8]. Idealized influence and inspirational motivation can be regarded as similar but distinct concepts to charismatic leadership because both are based on communication. Inspirational motivation is triggered when leaders express a vision and set goals. Leaders who show confidence, evoke moral conviction, and share values foster idealized influence. Like charismatic leadership signaling, these two dimensions seem well suited to reach remote workers. In contrast, intellectual stimulation (questioning and challenging followers) and individual consideration (attending to the need for growth) are harder to achieve in a remote setting. However, the concept of transformational leadership has been heavily criticized for its lack of a clear definition and causal evidence [2], [4], [5]. In the following, the focus is thus on studies providing clear, causal evidence based on observed follower behavior.

Charismatic leadership signaling and motivational talk lead to increased performance when used in face-to-face settings and can lead to better results than bonus payments. It has been shown that motivational talk and performance pay can reinforce each other and positively interact [11]. One study exposed temporary workers hired for a one-time job to either a neutral or a charismatic speech containing verbal and non-verbal charismatic signals such as rhetorical elements and body language [12]. The speeches were delivered face-to-face and the charismatic speech significantly increased the temporary workers’ performance compared to the neutral speech. Thus, these techniques work in one-time interactions.

However, the charismatic signals that can be sent in remote collaborations strongly depend on the chosen mode of communication, calling the effectiveness of such signals into question. The previously mentioned 2021 study [3] presents two experiments, both of which were conducted on Mturk. First, the authors investigated the impact of monetary contingent rewards and simple messages on performance, and then studied both instruments’ interaction. Second, they studied charismatic leadership signaling in the gig economy using a purely text-based method. All workers received written instructions about their job. As mentioned above, monetary contingent rewards increased output, but the relationship was not monotonic. In contrast, sending simple motivational messages reduced the delivered output, and there was no interaction between the performance pay and the motivational messages. To explain this backfiring, the researchers conducted the second experiment, where they exposed the workers to four different sets of instructions that build upon each other to implement charismatic signaling. The baseline setup involved neutral instructions containing a plain explanation of the task. The goal setup also included a goal that the workers should reach. In contrast, the full charisma setup contained the goal and a passage that aimed at providing frame, vision, and substance by using rhetorical elements. A fourth setup did not contain the goal part but the charisma part, to allow the researchers to disentangle the effect of goal setting from other charismatic signals. The results show that rhetorical signals indeed impacted the delivered performance (Figure 1). Even for this one-time, ten-minute job without any personal exchange and only text-based one-way communication, using the broad set of charismatic signals significantly increased output. Only announcing goals did not change the output, whereas using only a subset of charismatic signals backfired and even reduced delivered output compared to the neutral baseline. The charismatic signals only influenced the produced quantity but did not affect quality, which was high in all setups.

The above study focused on the impact of pure text-based communication, where non-verbal signals such as body gestures and tone of voice are missing [3]. The instructions built upon each other such that they all differed in length with the full charisma setup providing the longest instructions to the participants. In contrast, a different study from 2021 conducted several studies using video messages [2]. The sample mainly consisted of students hired to participate in an online experiment. The job was to prepare study cards for children in need. It should be noted that this job had a clear mission that might have provided intrinsic motivation in and of itself. In one setup, the students watched a video containing neutral instructions about the job they were expected to do. A trained actress used as many charismatic signals as possible in the other setup. Overall, the authors did not find a systematic impact of the two setups on delivered quality and quantity. The authors also studied the impact of culture by conducting the experiment in a range of different countries (Austria, France, India, and Mexico), but they found no cultural differences.

Another recent study from 2022 focuses on the impact of text-based, audio, and video messages on performance in the gig economy on Mturk [4]. The experiment included a typing task that had inherent meaning. Text fragments were to be digitalized to become part of an extensive collection that scholars and the general public could access. However, the task had no clear moral component, such as the tasks used in [2] and [12]. The gig workers were exposed either to a neutral task instruction or a task instruction using charismatic signals. The instructions were presented as plain text, a video, or an audio recording. In the video or audio recordings, a trained actor also used non-verbal charismatic signals such as facial expressions or intonation where appropriate. This setup allows the impacts of verbal and non-verbal signals in different communication modes to be disentangled. In addition, the data allow the impact of charismatic signaling in each communication mode to be studied. The results show that gig workers indeed perceived the leader to be more charismatic and prototypical if they used charismatic signals. This effect is most vital for video communication. The delivered output was significantly higher in the charisma video setup compared to the neutral video setup. In addition, the comparison of neutral text, versus audio, versus video messages revealed that sending a neutral video message backfired and led to lower output levels. In contrast, there were no significant differences in output when comparing the communication modes under the charisma setup. The delivered quality was not affected by the different setups and was generally high.

Summing up, results from the various studies reveal that the delivered quality of work on Mturk was high no matter which incentive system and which leadership signals were sent. This finding is very encouraging because even in an anonymous online labor market, most regular participants are not shirking. Given that the tasks outsourced to such labor markets are usually those that a machine or artificial intelligence cannot do, this is reassuring. The results differ between the studies concerning the impact of communication signals on the delivered output. However, taken together, it is possible to deduce the following: as a leader, it is essential to deliver a congruent picture. Using rhetorical elements helps to be perceived as charismatic. If non-verbal charismatic signals are absent, this has a strong negative effect. Also, if text-based communication only contains parts of the charismatic signaling instruments, this communication seems to be perceived as non-congruent, and the attempt to motivate followers backfires.