18/11/2015

Data journalism and ethics: What does the industry say?

 

Men's income in the United States has been falling for the past forty years...or so this Bloomberg article would have us believe. Yet, as Eric Portelance shows, this assertion was driven by misleading visualizations, such as insufficient data points and a truncated graph, that made superficial trends appear in uncorrelated data. Indeed, the census data employed by Bloomberg was not enough alone to drew any concrete conclusions about the historical story behind men's incomes.

To moral of this story is clear - all of the data we see in journalism is human driven and susceptible to the same kind of man made errors as traditional forms of reporting.

As one data journalist put it, "every algorithm, however it is written, contains human, and therefore editorial, judgment. The decisions made about what data to include and exclude add a layer of perspective to the information provided".

Misuse of statistics can occur easily - be it accidentally or to confirm an author's cognitive biases - and risks are increased when attempting to draw causation from large or unexplained datasets.

To help combat misinformation, and ensure that stories reflect best practices, this week The Ethical Journalism Network and the Donald W. Reynolds Journalism Institute at the University of Missouri launched the Accountable Journalism website - the world's largest database of media codes in a user-friendly and searchable Web application.

Users can search the database by topic, type of organization, country, region and date of creation. We took a look at the database's compilation of data journalism codes and complied a round up below.

The United Kingdom Committee on Publication Ethics Guidelines on Good Publication Practice

Founded in 1997, the Committee on Publication Ethics Guidelines on Good Publication Practice is a voluntary body assigned with developing good practice in the publication of science pieces. The code, last revised in 2011, contains a large section on data analysis standards.

Under the code, "data should be appropriately analysed, but inappropriate analysis does not necessarily amount to misconduct. Fabrication and falsification of data do constitute misconduct". To help authors meet this standard, the code provides an action mechanism framework to follow - it reads:

    "All sources and methods used to obtain and analyse data, including any electronic pre-processing, should be fully disclosed; detailed explanations should be provided for any exclusions.

    Methods of analysis must be explained in detail, and referenced, if they are not in common use.

    The post hoc analysis of subgroups is acceptable, as long as this is disclosed. Failure to disclose that the analysis was post hoc is unacceptable.

    The discussion section of a paper should mention any issues of bias which have been considered, and explain how they have been dealt with in the design and interpretation of the study."

The Swiss Declaration of Duties and Rights of Journalists 2000

This Declaration, utilized by the Swiss Press Council to make decisions, asserts that the "search for truth is at the heart of the act of informing. It presumes taking account of available and accessible data, respect for the integrity of documents (text, recording, image), verification and rectification." To this end, specific guidelines relating to survey conduct, freedom of information, covert research, and more, are outlined to ensure that best practices relating to data are outlined.

Code of Ethics of Lithuanian Journalists

Adopted by the Lithuanian Union of Journalists, the Lithuanian Association of Journalists, the Association of Publishers of Periodicals, the Lithuanian Radio and Television Association, the Lithuanian Radio and Television, and the Lithuanian Centre of Journalism on 25 March 1996, this code calls for "Truth, Honesty, Decency" in reporting. For data journalists specifically, the code states that data must only be used with "appropriate means of verification and evidence" and that the "author has to ensure that an opinion should be presented honestly and fairly, without any distortion of facts or data".

Mexico's El Universal's Code of Ethics

Journalists working for El Universal in Mexico, are guided by this code. In particular, under the honesty provision, data practices relating to surveys are subject to stringent guidelines:

"a) Information published in El Universal must distinguish clearly between investigations that use a scientific methodology  and studies that are not scientific, like interviews of people in the street, letters or phone calls, which must be reported in statistical terms. That needs to be done in a way that will be understandable by the reader.

    In using non scientific studies, one must clearly explain how they were made and what their limitations are. The nature of the selected sample and the margin of error must be published.

    Studies that do not reach a minimum level of scientific validity and reliability should not be reported in the language proper to scientific research.

    Great precaution should be taken in the use of scientific surveys to formulate important questions of public policies or to describe the popularity or rate of approval of public officials.

b) Tables and figures. Special care must be taken about the titles and graphic representation of scientific and non-scientific data so as to avoid confusion or wrong interpretation."

Learn more about industry codes and how they relate to data journalism on the Accountable Journalism website.

Photo: Justin Baeder

Comments