A Fractal Measure for Comparing the Work Effort of Human and Artificial Agents Performing Management Functions

Abstract

Abstract-Thanks to the growing sophistication of artificial agent technologies, businesses will increasingly face decisions of whether to have a human employee or artificial agent perform a particular function. This makes it desirable to have a common temporal measure for comparing the work effort that human beings and artificial agents can apply to a role. Existing temporal measures of work effort are formulated to apply either to human employees (e.g., FTE and billable hours) or computer-based systems (e.g., mean time to failure and availability) but not both. In this paper we propose a new temporal measure of work effort based on fractal dimension that applies equally to the work of human beings and artificial agents performing management functions. We then consider four potential cases to demonstrate the measure's diagnostic value in assessing strengths (e.g., flexibility) and risks (e.g., switch costs) reflected by the temporal work dynamics of particular managers. I. THE NEED FOR A COMMON TEMPORAL MEASURE OF WORK EFFORTHE increasing power and sophistication of artificial agent technology is allowing businesses to employ artificial agents in a growing number of roles. Artificial agents are no longer restricted simply to performing logistical functions such as resource scheduling, but are now capable of more complex interpersonal workplace behavior such as using social intelligence to effectively manage the limitations, abilities, and expectations of human employees [1], recognizing and manifesting culture-specific behaviors in interactions with human colleagues [2], and assessing the performance of human members of virtual teams [3]. It is thus gradually becoming more feasible to design artificial agents capable of performing the four key functions carried out by human managers, which are planning, organizing, leading, and controlling [4]. TAs a result of such recent and anticipated future advances, businesses will increasingly be faced with concrete decisions about whether, for example, the manager of a new corporate call center should be an experienced human manager or the latest artificial agent system. Such decisions will be shaped by a large number of strategic, financial, technological, political, legal, ethical, and operational factors. One particular element to be taken into account is that of temporal work effort: i.e., how much time would a human manager actually be able to dedicate to carrying out the necessary work functions, given the fact that physiological, cultural, legal, and ethical constraints limit the number of hours per week that a human being is capable of working? Similarly, how much time would an artificial agent be able to dedicate to carrying out the necessary work functions, given the fact that scheduled maintenance or unscheduled outages can limit the uptime of computer-based systems? Knowing how much time per day (or week, or other relevant time interval) a manager will be available to carry out his or her functions of planning, organizing, leading, and controlling becomes espe...

A tool for designing and evaluating the temporal work patterns of human and artificial agents

“…When assessing the work effort of artificial understood agents as computer-based systems, the relevant time intervals to be taken into consideration range from the macrotemporal level of several years (i.e., the lifespan of a typical commercial artificial agent system as utilized within a business) down to the microtemporal level of around 17 milliseconds (i.e., the time needed to generate a single screen refresh) or less [Gladden 2014]. As Gunther notes, there is no need to consider much smaller time intervals such as a single CPU cycle [Gunther 2005] that do not have an immediate relation to the larger-scale processes that we describe as an agent's meaningful "work.…”

“…When assessing the work effort of human managers, the relevant time-scales are similar; the time intervals to be taken into consideration range from the macrotemporal level of several years (i.e., the length of time that a human employee typically spends working for a particular organization) down to the microtemporal level of around 50 milliseconds (i.e., the length of time needed to hear or speak a single phoneme or to consciously perceive a single coherent experience or event [Gladden 2014].…”

“…Graphing an agent's work effort in two-dimensional space would be useful if the work effort displayed by a human or artificial agent manager at a particular instant of time were able to range across a continuous spectrum of values. However, in this case we have only a binary set of possible values: at any given instant, an agent is either focusing its attention on its work, or it is not [Marchetti 2000;Gladden 2014]. B.B.…”

“…For a set with high persistence (H ≈ 1), the point may be locked to a single x-value, reflecting a process with long memory. For a set with low persistence (H ≈ 0), the point "forgets" where it is and is free to move up and down the line segment, occupying many different x-values [Gladden 2014].…”

“…In developing new measures to compare the performance of human beings and artificial agents executing management functions, we find that one of the most crucial aspects of their work performance is the extent to which they effectively and efficiently use time. We have previously proposed a new temporal measure based on the Hurst exponent and fractal dimension that is useful for calculating and comparing the work effort dedicated to a given task within a particular time by either human personnel or artificial agent systems that are carrying out the four key management functions of planning, organizing, leading, or controlling [Gladden 2014]. In this paper, we further develop that proposed fractal measure as a means for identifying specific kinds of performance strengths and weaknesses that particular human or artificial agents are likely to demonstrate in such roles.…”