The obligation to justify and report results of systems such that harms can be mitigated.
Autonomous entities capable of interacting with the environment by receiving inputs, producing outputs, and performing operations; have some degree of control over their internal state which enables them to perform their operations autonomously; defined by the operations it performs.
Interactive and autonomous like actors, but also has the ability to adapt and change rules in order to perform its operations.
A property of an entity in the data model, singled out as relevant to its definition; a field in a schema; also known as a parameter depending on the context.
The ability for external stakeholders to assess the operations and quality standards of a system.
The frequency at which that data is processed. In Origo, four types:
Highest level of abstraction for a data model that captures business requirements and scopes projects.
A decision of resignation within an action space, either implicitly or explicitly.
A collection of related content types, or groups of information, found within an ICT, that is consumed by the end user.
An end user is able to interpret the results or have the results explained to them such that they can catch and resolve any errors.
One of the four basic functions of persistent storage; the addition of a new entry.
A lack of uniformity that makes a difference that can be perceived, measured, captured, and produced via an interaction; a precondition for experience.
A catalog of the fields in one or more schemas within a database system allowing an end user to navigate the database.
An abstraction of a domain, for purpose of translation into an ICT system, and thereby a set of distilled data structures representing real-world entities.
An abstraction of a domain, for purpose of translation into an ICT system, and thereby a set of distilled data structures representing objects in meatspace. For more details, view Model.
An attribute of data that tells the compiler how it is to be used. Included in Origo are:
One of the four basic functions of persistent storage; the deletion, deactivation, or removal of existing entries.
The "ought-to-be" statement involved in developing a solution to a wicked problem which is not based on scientific expertise but the moral and ethical attitude of the interventionist. Cannot be reconstructed by looking at the outcome of a solution as it is filigreed into the solution based on the underlying the judgment of the designer.
That there is one or more deontic premises for every solution for a wicked problem highlights the importance of involving multiple stakeholders in a transparent planning process.
A formal representation of an area of knowledge, influence, or activity; consists of real-world entities that eventually needs to be translated into software.
A form of misinformation achieved by spending a fraction of the resources that would be needed to tackle the ethical problems meant to be addressed.
The export of unethical research practices to countries where there are weaker legal and ethical frameworks and enforcing mechanisms.
Private actors attempting to self-regulate by lobbying against the introduction of legal norms and/or the weakening of enforcement such that compliance is limited.
The malpractice of doing increasingly less ethical work in a given context the lower the return of such ethical work in that context is perceived to be.
The malpractice of choosing, adapting, or revising ethical principles, guidelines, codes, frameworks, or other standards from a variety of available offers, in order to retrofit some pre-existing behaviors and justify them a posteriori instead of benchmarking against public, ethical standards.
A system that explains to a human the reasons behind a result. The human understands the system through its results; the human may not be able to interpret the inner workings of a model directly.
Vague beliefs without specific proof; can very be problematic when building product, especially if it is around a requirement.
An attribute of an object described in a schema. A set of field composes a record for an instance of an entity.
A population of actors, agents, and individuals that share resources to coordinate their activities effectively in order to achieve common goals.
Bias that exist in the dataset due unequal representation historically. This bias then perpetuate existing systemic problems into the future when used in a machine learning model.
Inconsistency in function that limits adoption by reducing the coherence of the operations performed by different actors, agents, and individuals within an organization.
When conceptual or technical challenges arise due to a poorly designed data model, pointing to how research into and translation of domain objects into the user interface was not well considered. May require costly refactoring.
Gaps in data collection which may not be effectively communicated to the data handler or unaccounted for in the processing, leading to biased results.
Inaccuracies in the data which may be due to the lack of technical rigor in the method in which it was collected.
Traditional entities that perform operations in an organization; autonomous, adaptive, interactive, and possess semanticisation capacity.
Something that makes an object discrete from the metaphysical flux around it, that gives it a coherent unity, a figure separated from ground.
A system that stores, retrieves, manipulates, transmits, and receives information (data) electronically, across networks, software, and hardware technologies.
A mechanism upon which objects or classes are based upon another object or class, retaining similar implementation or properties.
Information needed to accomplish a step in a task. May not always be in digital form as data, but the process of digitization usually refers to taking these inputs which may be found in different forms and creating a streamlined process where they can be digitized and moved through the system as discrete blendable data elements in the data model.
"... a tendency for an observer, first, to write or project or impose or inscribe a set of ontological assumptions onto a computational system (onto the system itself, onto the task domain, onto the relation between the two, and so forth), and then, second, to read those assumptions or their consequences back off the system, as if that constituted an independent empirical discovery or theoretical result."
The system is capable of being understood by humans. This is easier when the number of inputs is low.
Deliberate thought grounded in ethical commitment and responsible action that is appropriate to the situation in which it is deployed; human intelligence. [1]
Rittel claims that there are four kinds of judgments:
Sharing knowledge to solve problems; can be between industries, disciplines, generations, teams, etc. If knowledge is not transferred, it runs the risk of being lost or not meeting its full potential due to being trapped in a silo.
Detailed level of conceptual data model that contains all attributes and relationships needed for solution to problem.
A relationship where several instances of an object relates to several instances of another object.
Basic information about data to help users understand how to work with it; fields in a schema.
A place between the general and particular; a space of computation.
"...an intermediate realm between a proximal though ultimately ineffable connection, reminiscent of the familiar bumping and shoving of the world, and a more remote disconnection, a form of unabridgeable separation that lies at the root of abstraction and of the partial (and painful) subject-object divide."
A lack of situational information that may create faulty assumptions.
This word can get overloaded, but for the purposes of this glossary, it is a thing of interest out in meatspace that is distinct, unique, and self-contained. Through data modeling, these objects are translated into entities. See Object for more.
A collection of concepts embedded in a program that interfaces with a service or system.
The process of making one's judgment about what should be done explicit and communicating it to others so that there is a successful exchange of information about the foundations of the judgment, allowing it to be collectively discussed, reasoned, and/or argued over.
Whilst agreement may not be an outcome of objectification, the likelihood of one can be improved through collective learning. In the process of objectification, one also comes to a greater understanding of which questions are worthwhile to pursue, which have the greatest weight, and where there may be disagreements needing further testing and analysis.
Not to be confused with the process of making something objective, that is making a procedure that can be replicated by someone other than the creator. In contrast, objectification seeks to expose the judgment of the creator.
A relationship where an instance of an object relates to many instances of another object.
A relationship where an instance of an object relates to only one instance of another object.
The location from which data arrives as input. In Origo, this is classified as:
The owner of the data, where to own means to collect. If a change needs to be made, understanding who owns the data allows for understanding the scope of changes that can or cannot be made. Origo shows three types:
The specific, located, and singular; the form the particular takes depends on the perspective.
Information that can be directly linked to an individual's identity.
Implementation level of logical data model taking into account technical optimizations and constraints around data processing in the physical world within physical media.
Careless selection of data inputs for an algorithmic decision that may lead to biased results. Data might be missing in a way that disadvantages certain groups, or is structured at the wrong level of granularity.
Information that can be linked to other information to potentially reveal an individual's identity.
To try and anticipate the consequences of contemplated actions; thinking before acting.
Rittel lays out the following paradoxes of rationality:
These paradoxes move design away from a linear model of problem definition and problem solving (first generation system analysis) and positions it closer to experimental problem finding where solving the problem is the same as understanding it (second generation system analysis). That is, the solution space and the problem space are considered as one (wicked problem).
See wicked problem.
One of the four basic functions of persistent storage; the reading, reviewing, searching, or viewing of existing entries.
The powerful calculative rationality that display as intelligence possessed by machines; colloquially known as artificial intelligence.
Several fields in a row representing an instance of an entity for an object in meatspace.
The last step in a task. Necessary to identify first as part of bounding a task analysis. Can be identified by talking through the motivation of a task.
A type of actor; business process logic that is rigid and configured in detail.
The execution of a resource configured to be activated by an actor, agent, and/or individual.
To be free from harm or danger. Many questions of ethics are framed in terms of utilitarian trade-offs between risks and benefits, which is why safety is the primary protective criteria for disciplines with more mature ethical protocols.
The structure and design of a database. In a relational database, this is a set of tables and fields. The overall set of schemas driving an application composes the data model. Details in definition may vary slightly depending on database tooling (Oracle vs. Postgres, etc.)
The set of data inputs into a model is not representative of all stakeholder concerns resulting in biased outputs.
The process of semanticising information which may or may not be ethically neutral depending on information was collected, correlated, or interpreted.
Meaningful, veridical, comprehensible, accessible, and useful data that can be used for epistemic purposes and connected with a process of decision-making.
The operations performed by an agent to transform raw data into semantic information that is meaningful and truthful
A property of an entity generated through computational processing that describes how it has moved through the system; may or may not be able to be changed by a user. Success and error states are the minimum states generated through any process.
The basic unit of a task describing a change from the previous step. Step may be user or system generated.
Criteria which indicate when a solution has been found; does not exist in wicked problems where work only ends with lack of time, money, and/or patience.
Exists amongst stakeholders for a wicked problem given their local frame of the problem. Each stakeholder can only be an expert in guiding the process of handling the wicked problem because of their limited view; they cannot be subject matter experts of the problem itself.
A set of steps with a clear goal. Serves as a backbone to developing attributes, states, relationships, and other metadata that may be necessary to include in the schema for an entity.
A report summarizing a set of basic principles to guide the identification and resolution of ethical problems arising in research or involving ICT.
The degree of completeness to which users may access information, intentions, or behaviors, intentionally revealed through a process of disclosure, for use in decision-making; an enabling or impairing factor to doing ethics.
An abstraction of a domain, for purpose of translation into an ICT system, and thereby a set of distilled data structures representing real-world entities.
The structure and design of a database. In a relational database, this is a set of tables and fields. The overall set of schemas driving an application composes the data model. Details in definition may vary slightly depending on database tooling (Oracle vs. Postgres, etc.)
This word can get overloaded, but for the purposes of this glossary, it is a thing of interest out in meatspace that is distinct, unique, and self-contained. Through data modeling, these objects are translated into entities. See Object for more.
A catalog of the fields in one or more schemas within a database system allowing an end user to navigate the database.
A data schema or structure that corresponds to an object in meatspace.
An example of an object.
Several fields in a row representing an instance of an entity for an object in meatspace.
An attribute of an object described in a schema. A set of field composes a record for an instance of an entity.
A property of an entity in the data model, singled out as relevant to its definition; a field in a schema; also known as a parameter depending on the context.
Basic information about data to help users understand how to work with it; fields in a schema.
A property of an entity generated through computational processing that describes how it has moved through the system; may or may not be able to be changed by a user. Success and error states are the minimum states generated through any process.
A connection between two entities. For more details, see Relationship.
A set of steps with a clear goal. Serves as a backbone to developing attributes, states, relationships, and other metadata that may be necessary to include in the schema for an entity.
The basic unit of a task describing a change from the previous step. Step may be user or system generated.
The first step of a task. Can be identified by talking through the motivation of a task. Necessary to identify upfront when doing a task analysis.
The last step in a task. Necessary to identify first as part of bounding a task analysis. Can be identified by talking through the motivation of a task.
Information needed to accomplish a step in a task. May not always be in digital form as data, but the process of digitization usually refers to taking these inputs which may be found in different forms and creating a streamlined process where they can be digitized and moved through the system as discrete blendable data elements in the data model.
A lack of uniformity that makes a difference that can be perceived, measured, captured, and produced via an interaction; a precondition for experience.
An attribute of data that tells the compiler how it is to be used. Included in Origo are:
The frequency at which that data is processed. In Origo, four types:
The location from which data arrives as input. In Origo, this is classified as:
The owner of the data, where to own means to collect. If a change needs to be made, understanding who owns the data allows for understanding the scope of changes that can or cannot be made. Origo shows three types:
Information that can be directly linked to an individual's identity.
Information that can be linked to other information to potentially reveal an individual's identity.
A mechanism upon which objects or classes are based upon another object or class, retaining similar implementation or properties.
A counting of how many things can relate to something else.
A relationship where an instance of an object relates to only one instance of another object.
A relationship where an instance of an object relates to many instances of another object.
A relationship where several instances of an object relates to several instances of another object.
The process of making one's judgment about what should be done explicit and communicating it to others so that there is a successful exchange of information about the foundations of the judgment, allowing it to be collectively discussed, reasoned, and/or argued over.
Whilst agreement may not be an outcome of objectification, the likelihood of one can be improved through collective learning. In the process of objectification, one also comes to a greater understanding of which questions are worthwhile to pursue, which have the greatest weight, and where there may be disagreements needing further testing and analysis.
Not to be confused with the process of making something objective, that is making a procedure that can be replicated by someone other than the creator. In contrast, objectification seeks to expose the judgment of the creator.
The "ought-to-be" statement involved in developing a solution to a wicked problem which is not based on scientific expertise but the moral and ethical attitude of the interventionist. Cannot be reconstructed by looking at the outcome of a solution as it is filigreed into the solution based on the underlying the judgment of the designer.
That there is one or more deontic premises for every solution for a wicked problem highlights the importance of involving multiple stakeholders in a transparent planning process.
Exists amongst stakeholders for a wicked problem given their local frame of the problem. Each stakeholder can only be an expert in guiding the process of handling the wicked problem because of their limited view; they cannot be subject matter experts of the problem itself.
A decision of resignation within an action space, either implicitly or explicitly.
Something that makes an object discrete from the metaphysical flux around it, that gives it a coherent unity, a figure separated from ground.
An abstraction of a domain, for purpose of translation into an ICT system, and thereby a set of distilled data structures representing objects in meatspace. For more details, view Model.
A record of an event that has taken pace within an ICT.
The execution of a resource configured to be activated by an actor, agent, and/or individual.
One of the four basic functions of persistent storage; the update or editing of existing entries.
One of the four basic functions of persistent storage; the deletion, deactivation, or removal of existing entries.
One of the four basic functions of persistent storage; the reading, reviewing, searching, or viewing of existing entries.
One of the four basic functions of persistent storage; the addition of a new entry.
Inconsistency in function that limits adoption by reducing the coherence of the operations performed by different actors, agents, and individuals within an organization.
A population of actors, agents, and individuals that share resources to coordinate their activities effectively in order to achieve common goals.
A report summarizing a set of basic principles to guide the identification and resolution of ethical problems arising in research or involving ICT.
A system that stores, retrieves, manipulates, transmits, and receives information (data) electronically, across networks, software, and hardware technologies.
The process of semanticising information which may or may not be ethically neutral depending on information was collected, correlated, or interpreted.
The operations performed by an agent to transform raw data into semantic information that is meaningful and truthful
Meaningful, veridical, comprehensible, accessible, and useful data that can be used for epistemic purposes and connected with a process of decision-making.
A place between the general and particular; a space of computation.
"...an intermediate realm between a proximal though ultimately ineffable connection, reminiscent of the familiar bumping and shoving of the world, and a more remote disconnection, a form of unabridgeable separation that lies at the root of abstraction and of the partial (and painful) subject-object divide."
The specific, located, and singular; the form the particular takes depends on the perspective.
The powerful calculative rationality that display as intelligence possessed by machines; colloquially known as artificial intelligence.
Deliberate thought grounded in ethical commitment and responsible action that is appropriate to the situation in which it is deployed; human intelligence. [1]
Rittel claims that there are four kinds of judgments:
Criteria which indicate when a solution has been found; does not exist in wicked problems where work only ends with lack of time, money, and/or patience.
A different class of problems from those handled by engineering and/or science; societal problems which are ill-defined and which require a political position in order to resolve as no solution can actually exist. Each problem is unique, and each solution is a one-shot operation. Wicked problems cannot be simulated in a lab.
"The process of formulating a problem and of conceiving a solution are identical, since every specification of the problem is a specification of the direction in which a treatment is considered... one cannot understand the problem without knowing its context; one cannot meaningfully search for information without the orientation of a solution concept; one cannot first understand then solve." [1]
"A problem can be stated as a discrepancy, as something as it is compared with something as it ought to be... you look for reasons for the existence of this discrepancy, the cause and the explanation. And the trouble is that in wicked problems there are many explanations for the same discrepancy and there is no test for which of these explanations is the best one." [2]
Origo treats ethics as a wicked problem where the only way to move towards good is by improving the time to understanding the underlying design problem, which is, under this framework, the same as improving the chances of creating a more ethical design solution.
Implementation level of logical data model taking into account technical optimizations and constraints around data processing in the physical world within physical media.
Detailed level of conceptual data model that contains all attributes and relationships needed for solution to problem.
Highest level of abstraction for a data model that captures business requirements and scopes projects.
A formal representation of an area of knowledge, influence, or activity; consists of real-world entities that eventually needs to be translated into software.
A collection of concepts embedded in a program that interfaces with a service or system.
A collection of related content types, or groups of information, found within an ICT, that is consumed by the end user.
Interactive and autonomous like actors, but also has the ability to adapt and change rules in order to perform its operations.
Traditional entities that perform operations in an organization; autonomous, adaptive, interactive, and possess semanticisation capacity.
Autonomous entities capable of interacting with the environment by receiving inputs, producing outputs, and performing operations; have some degree of control over their internal state which enables them to perform their operations autonomously; defined by the operations it performs.
A type of actor; business process logic that is rigid and configured in detail.
A system that stores, retrieves, manipulates, transmits, and receives information (data) electronically, across networks, software, and hardware technologies.
A report summarizing a set of basic principles to guide the identification and resolution of ethical problems arising in research or involving ICT.
A population of actors, agents, and individuals that share resources to coordinate their activities effectively in order to achieve common goals.
Inconsistency in function that limits adoption by reducing the coherence of the operations performed by different actors, agents, and individuals within an organization.
Exists amongst stakeholders for a wicked problem given their local frame of the problem. Each stakeholder can only be an expert in guiding the process of handling the wicked problem because of their limited view; they cannot be subject matter experts of the problem itself.
Autonomous entities capable of interacting with the environment by receiving inputs, producing outputs, and performing operations; have some degree of control over their internal state which enables them to perform their operations autonomously; defined by the operations it performs.
Traditional entities that perform operations in an organization; autonomous, adaptive, interactive, and possess semanticisation capacity.
Interactive and autonomous like actors, but also has the ability to adapt and change rules in order to perform its operations.
A type of actor; business process logic that is rigid and configured in detail.
A decision of resignation within an action space, either implicitly or explicitly.
A formal representation of an area of knowledge, influence, or activity; consists of real-world entities that eventually needs to be translated into software.
An abstraction of a domain, for purpose of translation into an ICT system, and thereby a set of distilled data structures representing objects in meatspace. For more details, view Model.
A collection of related content types, or groups of information, found within an ICT, that is consumed by the end user.
A collection of concepts embedded in a program that interfaces with a service or system.
Highest level of abstraction for a data model that captures business requirements and scopes projects.
Detailed level of conceptual data model that contains all attributes and relationships needed for solution to problem.
Implementation level of logical data model taking into account technical optimizations and constraints around data processing in the physical world within physical media.
A different class of problems from those handled by engineering and/or science; societal problems which are ill-defined and which require a political position in order to resolve as no solution can actually exist. Each problem is unique, and each solution is a one-shot operation. Wicked problems cannot be simulated in a lab.
"The process of formulating a problem and of conceiving a solution are identical, since every specification of the problem is a specification of the direction in which a treatment is considered... one cannot understand the problem without knowing its context; one cannot meaningfully search for information without the orientation of a solution concept; one cannot first understand then solve." [1]
"A problem can be stated as a discrepancy, as something as it is compared with something as it ought to be... you look for reasons for the existence of this discrepancy, the cause and the explanation. And the trouble is that in wicked problems there are many explanations for the same discrepancy and there is no test for which of these explanations is the best one." [2]
Origo treats ethics as a wicked problem where the only way to move towards good is by improving the time to understanding the underlying design problem, which is, under this framework, the same as improving the chances of creating a more ethical design solution.
Meaningful, veridical, comprehensible, accessible, and useful data that can be used for epistemic purposes and connected with a process of decision-making.
The operations performed by an agent to transform raw data into semantic information that is meaningful and truthful
The process of semanticising information which may or may not be ethically neutral depending on information was collected, correlated, or interpreted.
A place between the general and particular; a space of computation.
"...an intermediate realm between a proximal though ultimately ineffable connection, reminiscent of the familiar bumping and shoving of the world, and a more remote disconnection, a form of unabridgeable separation that lies at the root of abstraction and of the partial (and painful) subject-object divide."
The specific, located, and singular; the form the particular takes depends on the perspective.
Something that makes an object discrete from the metaphysical flux around it, that gives it a coherent unity, a figure separated from ground.
Deliberate thought grounded in ethical commitment and responsible action that is appropriate to the situation in which it is deployed; human intelligence. [1]
Rittel claims that there are four kinds of judgments:
The powerful calculative rationality that display as intelligence possessed by machines; colloquially known as artificial intelligence.
The "ought-to-be" statement involved in developing a solution to a wicked problem which is not based on scientific expertise but the moral and ethical attitude of the interventionist. Cannot be reconstructed by looking at the outcome of a solution as it is filigreed into the solution based on the underlying the judgment of the designer.
That there is one or more deontic premises for every solution for a wicked problem highlights the importance of involving multiple stakeholders in a transparent planning process.
The process of making one's judgment about what should be done explicit and communicating it to others so that there is a successful exchange of information about the foundations of the judgment, allowing it to be collectively discussed, reasoned, and/or argued over.
Whilst agreement may not be an outcome of objectification, the likelihood of one can be improved through collective learning. In the process of objectification, one also comes to a greater understanding of which questions are worthwhile to pursue, which have the greatest weight, and where there may be disagreements needing further testing and analysis.
Not to be confused with the process of making something objective, that is making a procedure that can be replicated by someone other than the creator. In contrast, objectification seeks to expose the judgment of the creator.
Criteria which indicate when a solution has been found; does not exist in wicked problems where work only ends with lack of time, money, and/or patience.
To try and anticipate the consequences of contemplated actions; thinking before acting.
Rittel lays out the following paradoxes of rationality:
These paradoxes move design away from a linear model of problem definition and problem solving (first generation system analysis) and positions it closer to experimental problem finding where solving the problem is the same as understanding it (second generation system analysis). That is, the solution space and the problem space are considered as one (wicked problem).
See wicked problem.
A lack of situational information that may create faulty assumptions.
Vague beliefs without specific proof; can very be problematic when building product, especially if it is around a requirement.
"... a tendency for an observer, first, to write or project or impose or inscribe a set of ontological assumptions onto a computational system (onto the system itself, onto the task domain, onto the relation between the two, and so forth), and then, second, to read those assumptions or their consequences back off the system, as if that constituted an independent empirical discovery or theoretical result."
When conceptual or technical challenges arise due to a poorly designed data model, pointing to how research into and translation of domain objects into the user interface was not well considered. May require costly refactoring.
Sharing knowledge to solve problems; can be between industries, disciplines, generations, teams, etc. If knowledge is not transferred, it runs the risk of being lost or not meeting its full potential due to being trapped in a silo.
An end user is able to interpret the results or have the results explained to them such that they can catch and resolve any errors.
The health, happiness, and well-being of a group that should be protected.
To be free from harm or danger. Many questions of ethics are framed in terms of utilitarian trade-offs between risks and benefits, which is why safety is the primary protective criteria for disciplines with more mature ethical protocols.
The obligation to justify and report results of systems such that harms can be mitigated.
The ability for external stakeholders to assess the operations and quality standards of a system.
The system is capable of being understood by humans. This is easier when the number of inputs is low.
A system that explains to a human the reasons behind a result. The human understands the system through its results; the human may not be able to interpret the inner workings of a model directly.
The degree of completeness to which users may access information, intentions, or behaviors, intentionally revealed through a process of disclosure, for use in decision-making; an enabling or impairing factor to doing ethics.
The malpractice of doing increasingly less ethical work in a given context the lower the return of such ethical work in that context is perceived to be.
The export of unethical research practices to countries where there are weaker legal and ethical frameworks and enforcing mechanisms.
Private actors attempting to self-regulate by lobbying against the introduction of legal norms and/or the weakening of enforcement such that compliance is limited.
A form of misinformation achieved by spending a fraction of the resources that would be needed to tackle the ethical problems meant to be addressed.
The malpractice of choosing, adapting, or revising ethical principles, guidelines, codes, frameworks, or other standards from a variety of available offers, in order to retrofit some pre-existing behaviors and justify them a posteriori instead of benchmarking against public, ethical standards.
Careless selection of data inputs for an algorithmic decision that may lead to biased results. Data might be missing in a way that disadvantages certain groups, or is structured at the wrong level of granularity.
Bias that exist in the dataset due unequal representation historically. This bias then perpetuate existing systemic problems into the future when used in a machine learning model.
The set of data inputs into a model is not representative of all stakeholder concerns resulting in biased outputs.
Gaps in data collection which may not be effectively communicated to the data handler or unaccounted for in the processing, leading to biased results.
Inaccuracies in the data which may be due to the lack of technical rigor in the method in which it was collected.
Inaccuracies in the data which may be due to the lack of technical rigor in the method in which it was collected.
Gaps in data collection which may not be effectively communicated to the data handler or unaccounted for in the processing, leading to biased results.
The set of data inputs into a model is not representative of all stakeholder concerns resulting in biased outputs.
Bias that exist in the dataset due unequal representation historically. This bias then perpetuate existing systemic problems into the future when used in a machine learning model.
Careless selection of data inputs for an algorithmic decision that may lead to biased results. Data might be missing in a way that disadvantages certain groups, or is structured at the wrong level of granularity.
Data is not being updated at a cadence that is sensible for the use case.
Sharing knowledge to solve problems; can be between industries, disciplines, generations, teams, etc. If knowledge is not transferred, it runs the risk of being lost or not meeting its full potential due to being trapped in a silo.
When conceptual or technical challenges arise due to a poorly designed data model, pointing to how research into and translation of domain objects into the user interface was not well considered. May require costly refactoring.
"... a tendency for an observer, first, to write or project or impose or inscribe a set of ontological assumptions onto a computational system (onto the system itself, onto the task domain, onto the relation between the two, and so forth), and then, second, to read those assumptions or their consequences back off the system, as if that constituted an independent empirical discovery or theoretical result."
Vague beliefs without specific proof; can very be problematic when building product, especially if it is around a requirement.
A lack of situational information that may create faulty assumptions.
The malpractice of choosing, adapting, or revising ethical principles, guidelines, codes, frameworks, or other standards from a variety of available offers, in order to retrofit some pre-existing behaviors and justify them a posteriori instead of benchmarking against public, ethical standards.
A form of misinformation achieved by spending a fraction of the resources that would be needed to tackle the ethical problems meant to be addressed.
Private actors attempting to self-regulate by lobbying against the introduction of legal norms and/or the weakening of enforcement such that compliance is limited.
The export of unethical research practices to countries where there are weaker legal and ethical frameworks and enforcing mechanisms.
The malpractice of doing increasingly less ethical work in a given context the lower the return of such ethical work in that context is perceived to be.
To try and anticipate the consequences of contemplated actions; thinking before acting.
Rittel lays out the following paradoxes of rationality:
These paradoxes move design away from a linear model of problem definition and problem solving (first generation system analysis) and positions it closer to experimental problem finding where solving the problem is the same as understanding it (second generation system analysis). That is, the solution space and the problem space are considered as one (wicked problem).
See wicked problem.
The degree of completeness to which users may access information, intentions, or behaviors, intentionally revealed through a process of disclosure, for use in decision-making; an enabling or impairing factor to doing ethics.
A system that explains to a human the reasons behind a result. The human understands the system through its results; the human may not be able to interpret the inner workings of a model directly.
The system is capable of being understood by humans. This is easier when the number of inputs is low.
The ability for external stakeholders to assess the operations and quality standards of a system.
The obligation to justify and report results of systems such that harms can be mitigated.
To be free from harm or danger. Many questions of ethics are framed in terms of utilitarian trade-offs between risks and benefits, which is why safety is the primary protective criteria for disciplines with more mature ethical protocols.
The health, happiness, and well-being of a group that should be protected.
An end user is able to interpret the results or have the results explained to them such that they can catch and resolve any errors.