Evidence-Informed Lawmaking and the Use of Evidence by the European Parliament: The Case of Harmful Content in the Digital Services Act

Table of Contents: 1. Introduction. – 1.1. Better lawmaking as a joint effort. – 1.2. Methodology. – 2. Shortcomings in the Commission impact assessment. – 2.1 Regulating platforms’ standard terms. – 2.2. Regulating recommender systems. – 2.3 Evidence, strengths and weaknesses in the impact assessment. – 3. European Parliament’s amendments. – 3.1. Amendments to Commission draft Article 12 on platforms’ standard terms. – 3.2. Amendments to Commission draft Article 29 on recommender systems. – 3.3. The use of evidence in the European Parliament regarding harmful content in the DSA. – 4. General observations on evidence-informed EU lawmaking. – 5. Implications for reform.

Abstract: EU evidence-informed lawmaking is criticised as European Parliament and the Council frequently introduce substantial amendments without conducting impact assessments. This criticism overlooks that there is little research on evidence actually used by the European Parliament. An analysis of Parliament’s amendments to the provisions on content moderation and recommender systems in the DSA shows that Parliament makes much more use of sources on fundamental rights than the European Commission, and generally lives up to its commitment to evidence-informed lawmaking. Its amendments to Article 14 and 27 DSA may compensate for some of the shortcomings of the Commission impact assessment to the DSA – a superficial human rights impact assessment, lack of attention for the role of private actors and focus on terms and conditions. More generally, the amendments to the DSA demonstrate the difficulty to conduct additional impact assessments for substantial amendments, in line with the Interinstitutional Agreement on Better Lawmaking, and showcase that evidence-informed lawmaking does not take away the need to make political choices. Instead, the drafting process suggests that the use of relevant, authoritative evidence does not necessarily mean that amendments will be adopted, nor has it consistently resulted in amendments that are of high legislative quality. The article concludes with five suggestions for the upcoming reform of better regulation.

Keywords: Better Regulation – evidence – impact assessments – Digital Services Act – harmful content – recommender systems.

1.   Introduction

1.1.  Improving evidence-informed lawmaking: the example of the DSA

Evidence-informed lawmaking in the EU remains incomplete, as the European Parliament does not systematically conduct impact assessments for substantial amendments to legislative proposals.[1] The Draghi report, the Letta report, and the OECD recommend that the European Parliament and Council comply with its own commitment to evidence-informed lawmaking, to ensure that EU law is based on the best available evidence.[2] The Commission has since published a call for evidence to revise the Better Regulation Guidelines, in line with the recommendations in the Draghi report and the Letta Report.[3]

However, while the European Parliament has committed itself to conduct such assessments for significant amendments, it is not clear which amendments are significant and therefore require impact assessments.[4] A binding obligation might also be detrimental to institutional balance.[5] What’s more, there is little information on the evidence actually used by the European Parliament, nor does systematic scrutiny exist on what happens with evidence after a Commission legislative proposal is discussed and amended in Parliament.

This Article presents a different perspective on the potential of the European Parliament to support evidence-informed lawmaking. A binding obligation to conduct impact assessment might not only have a chilling effect on proposing such amendments, even if necessary – it would also overlook that it may be difficult to estimate the impact of proposed amendments that may well contradict one another. That does not mean Parliament should not benefit from evidence in its decision-making. To the contrary, Parliament seems well placed to support evidence-informed legislation.[6] Indeed, the potential role of Parliament to ensure that lawmaking is informed by relevant, robust evidence is considerable. For example, a poor selection of policy options may be mitigated if the European Parliament and Council critically question the evidence supporting impact assessments and suggest alternative options. An area where such potential becomes apparent is EU consumer law, where the benefits of evidence-informed lawmaking are not fully realized. In this area of law, Commission impact assessments, essential for analysing the future impact of EU legislative proposals, are criticized as incomplete,[7] their one-sided use of evidence,[8] or for justifying choices already made.[9] Shortcomings are also visible in the impact assessment to the Digital Services Act (hereafter: DSA).[10]

The European Parliament collected extensive sources of evidence in preparation of the DSA. A considerable number of Parliamentary committees issued opinions and reports on the DSA, and it was preceded by three own-initiative reports,[11] building on three external assessments of the legal framework,[12] and various internal and external studies.[13] The debate in Parliament further had the possibility to benefit from the appraisal of the Commission impact assessment,[14] the EDPS opinion,[15] as well as studies commissioned by the European Parliament,[16] and a workshop with experts.[17]

The Commission’s impact assessment showed shortcomings – for example, it did not justify some of the regulatory choices regarding harmful content, and its assessments of impact on fundamental rights is superficial.[18] Parliament was well-placed to mitigate these shortcomings, considering the extensive amount of evidence that was collected prior to and after the adoption of the draft DSA by the Commission. This article examines how the evidence collected by or for the European Parliament, in its amendments to the DSA regarding the dissemination of harmful content could compensate for the shortcomings in the Commission’s impact assessment to the DSA, if at all?

1.2.  Methodology

This Article will focus on the legislator’s choices in addressing harmful yet legal content in the preparation of the DSA as a case study for the potential role of the European Parliament in evidence-informed lawmaking. The DSA is vital for protecting consumers in a digitalized platform economy, even though it aims to address illegal content rather than promote consumer protection.[19] It forms part of an area of law, EU consumer law, where impact assessments have already been criticized repeatedly,[20] and the potential of Parliament to mitigate these shortcomings may be considerable. In addition, the DSA was chosen as a case study because it was one of the key political priorities of the European Commission, the European Parliament and the Council,[21] and therefore likely to generate considerable political interest and evidence to support political debate.

Within the DSA, harmful content presents a particular regulatory challenge for the EU legislator. The European legislator has addressed harmful yet legal content and its dissemination through Article 14 on terms and conditions, and Articles 27 and 38 DSA on recommender systems. However, initially, despite the availability of extensive evidence, there was disagreement within Parliament whether the DSA should also include harmful content.[22] For one, harmful content is notoriously difficult to define. What content should be harmful, and why?[23] Harmful content has many forms,[24] and is not necessarily illegal, and therefore not subject to the same removal obligations as illegal content.[25] Further, limiting individuals’ freedom to post content online, even if offensive or ill-informed, may for example be difficult to reconcile with users’ freedom of expression, a prerequisite for a democratic society,[26] that includes the freedom to ‘shock, offend and disturb’,[27] which often takes place online.[28] Further, harmful content may serve a clear, legitimate, purpose – content that depicts nudity or violence portraying historical events may serve educational purposes, or defamatory statements may attempt to draw attention to racism.[29] However, not removing harmful content may inhibit the rights of other users – for example, children’s rights, if they are confronted with graphic or violent content. Not removing harmful content may aggravate discrimination and undermine human dignity, especially if enforcement targets marginalized groups while indulging powerful groups and public figures.[30] To meet this regulatory challenge, the European legislator not only has to be sufficiently informed, but also needs to carefully balance rights and interests of different parties. Thus, the European Parliament was well placed to compensate for shortcomings in the impact assessment for the DSA, but did it do so?

As the DSA was a key political priority and generated considerable political interest, the drafting process may differ from other measures that were not as politically significant. Possibly, for other measures, Parliament may collect fewer amount of sources. Taking this into account, the article will reflect on the findings more widely, particularly in paragraphs 4 and 5, drawing from reports on EU impact assessments and the approach of EU Parliament, as well as literature on IAs, not only in EU consumer law but more generally.

Analysing the use of evidence in the European Parliament is difficult, as it is unclear which evidence Parliament has used in support of (proposed) amendments. The justifications to proposed amendments are short and rarely refer to sources.[31] Therefore, verifying whether amendments have considered the data and arguments presented in the impact assessment, and compensate for weaknesses therein, is problematic. Furthermore, the discussions preceding amendments within political fractions are not public, but some political parties represented in the lead and associate committees have published their views on the Commission draft DSA,[32] and the EFA/Greens even provided a helpful analysis on its influence in drafting the DSA,[33] having previously conducted a survey.[34]

In order to discover possible sources of amendments, the article will compare proposed amendments and the short justifications to various sources: expert briefings before the European Parliament and sources therein, position papers from stakeholders that have met with the rapporteurs or members of associate committees,[35] including representatives of Member States that were in the process of drafting, or had adopted various national measures targeting harmful content,[36] as well as stakeholders who responded to Article 12 and 29 Commission draft DSA[37] and Parliament’s position.[38] The paper will further compare the content of and evidence used in amendments, if any, with evidence used by the European Commission in preparing the impact assessment, and other potentially relevant sources of evidence, ranging from academic publications, European Union initiatives to regulate harmful content, ECHR case law, Council of Europe Recommendations and positions adopted by relevant enforcement authorities.

The provisions on harmful content are linked to Articles 34 (risk assessment) and 35 (mitigation of risks) that expressly require considering the impact of terms and conditions and recommender systems on systemic risk and taking measures to mitigate said risks. It is however beyond the scope of this article to extensively analyse (the drafting of) Articles 34 and 35 DSA.[39] It is similarly beyond the scope of this Article to consider the relationship between the EU, the Charter, and the ECHR.[40]

Section 2 will assess the shortcomings in the impact assessment, and Section 3 will explore whether the European Parliament’s amendments and the evidence used therein might compensate for these weaknesses. Section 4 will answer the research question and make some general observations on evidence-informed EU lawmaking, especially for EU consumer law. Section 5 will consider implications for the upcoming revision of the Better Regulation Guidelines.

2.   Shortcomings in the Commission impact assessment

The impact assessment considered three policy options to address the problem drivers identified in the impact assessment.[41] The impact assessment argues that regulating recommender systems would both help address systemic risks, correct information asymmetries and empower consumers.[42] Recommender systems may play an important role in, for example, shaping public debate, either by encouraging the inclusion of more diverse perspectives or by narrowing debate and hindering the visibility of, for example, minority views.[43] The regulation of terms and conditions is not separately discussed in the impact assessment, which seems more focused on preventing the dissemination of harmful content than addressing individual harmful content, which is left to platforms’ terms and conditions.

2.1.  Regulating platforms’ standard terms

Content moderation had been widely criticized as untransparent, difficult to reconcile with users’ fundamental rights, including their freedom of expression[44] and access to an independent court.[45] Accordingly, the impact assessment identified as one of the problem drivers that private actors – platforms – make decisions – for example in moderating content – that have significant impact on users’ rights.[46] Thus, Article 12(1) Commission draft DSA obliged platforms to communicate the restrictions on the use of their services ‘in clear and unambiguous language’, and to ‘act in a diligent, objective and proportionate manner in applying and enforcing’ these restrictions. Article 12(2) Commission draft DSA adds that platforms should have ‘due regard to the rights and legitimate interests of all parties involved’, referring to the Charter. This may also mean that harmful content stays online.

Two weaknesses become apparent within this part of the impact assessment – i) the lack of debate on regulatory choices, and ii) its superficial analysis of human rights.

2.1.1.   Shortcoming 1: Lack of debate on regulatory choices

The choice to continue allowing for the removal of harmful content, while providing criteria for the terms on the basis of which harmful content is removed, is to be welcomed, but the discussion whether this should take place through legislation or private regulation is missing. The choice to regulate through private regulation, in the form of terms and conditions, is not illogical. Platforms are, after all, the actors currently making decisions impacting users’ fundamental rights, and are best placed to enforce their own contractual terms on removing harmful content. Their role is also in line with the principle of the freedom of contract.[47] This may mean that platforms’ terms go beyond what is required by the DSA,[48] and even that the competence of platforms to determine their terms and conditions on the basis of Article 12 Commission draft DSA could stand in the way of judicial review of those terms.[49] While Article 12 could indeed have been clearer, it is not evident that it would have stood in the way of judicial screening of platforms standard terms. After all, the Commission underlines that the DSA ‘will be complementary to the consumer protection acquis’, specifically referring to the Unfair Terms Directive.[50]

The use of self-regulation rather than state (or EU) regulation had been defended in order to protect freedom of speech, as targeting harmful content via legislation may result in legal removal obligations for harmful content,[51] which is likely to have a chilling effect on the freedom of speech.[52]

In the case of the DSA, though, the discussion whether harmful content should be regulated via legislation or private regulation is entirely absent. This is problematic because legislation should at least be considered. The requirement in Article 52(1) Charter that all limitations on the rights recognized in the Charter must take place via law indicates the need for legislation. The development of legislation is also preferable from the perspective of the rule of law – the development of EU legislation is, after all, surrounded by more safeguards, including respect for human rights, than private regulation. For harmful content, concerns over over-removal could have also been a reason to limit platforms’ roles.[53] The EU in addition is not unfamiliar with this possibility, as both legislation and alternative regulation have been used for regulating harmful content.[54]

Consequently, the impact assessment does not recognize the impact of leaving the moderation of harmful content largely to online platforms on users’ fundamental rights. Essentially, the definition of ‘harmful’ is left to platforms, who, at the time of drafting the DSA, were criticized for their vague terms and conditions.[55] Consequently, the problem signalled by the impact assessment remains.[56]

2.1.2. Shortcoming 2: Superficial assessment of fundamental rights

State (EU) actors have a positive obligation to ensure that third parties – such as platforms – do not unduly restrict the exercise of fundamental rights such as the freedom of speech.[57] Therefore, state actors must develop foreseeable rules allowing the use of online information without fear of repercussions, especially for journalists.[58] The Better Regulation Guidelines acknowledge the obligation to consider the impact of initiatives on fundamental rights and to comply with the Charter.[59]

The impact assessment outlines the ways in which the intervention seeks to mitigate online censorship. For one, the choice of focus should reduce undue limitations to fundamental rights.[60] Particularly, undue limitations should be avoided by the focus on illegal rather than harmful yet legal content, and the choice to ban prior authorization schemes, as well as general monitoring obligations. Indeed, the ECHR has similarly distinguished between legal and illegal content and especially content that is clearly illegal should be removed, also in view of the interests and rights of others and the general public.[61] Banning prior authorization schemes is also in line with ECHR rulings that emphasized that, while permissible, this option carries significant risk, which requires especially careful scrutiny of the need for such schemes, ‘justified only in exceptional circumstances’.[62] In addition, the draft DSA would help protect freedom of expression by introducing safeguards in the form of complaint systems and transparency requirements. Lastly, the draft DSA would help reduce an unsafe illegal environment, which may have a chilling effect on the freedom of speech.[63]

However, the reasoning of the impact assessment fails to explore the conditions for limiting fundamental rights in line with ECHR case law.[64] Platforms’ standard terms that form the basis for removal are vague and do not always provide users with a clear, let alone legitimate reason for removal, contrary to ECHR requirements, that limitations to the freedom of expression are foreseeable, accessible and predictable.[65] The fundamental rights assessment in the impact assessment also remains superficial in other aspects. For example, while the impact assessment considers illegal content, it does not distinguish between illegal content and clearly or manifestly illegal content that does not require platforms, rather than courts, to determine the legality of content. Additionally, the impact assessment does not clarify if fundamental rights instruments, particularly the Charter, will be directly applicable between private parties, even though the impact assessment considered private parties’ impact on other private parties’ fundamental rights one of the problem drivers.

The impact assessment also does not consider how parties’ rights should be balanced, even though the CJEU and the ECHR have repeatedly recognized the need to balance the freedom to express oneself online with the general interest[66] and the protection of others’ rights, such as privacy, the protection of personal data, the rights of the child, or the right to conduct a business.[67] Particularly, the CJEU has ruled that protection of personal data may generally prevail over the freedom of information also covered under Article 11 Charter, taking into account the nature and sensitivity of the information, the impact of publication thereof on a person’s private life, and the interest of the public in having that information available.[68]

2.2   Regulating recommender systems

The impact assessment identifies the design of platforms and its untransparent use of recommender systems, as relevant for emerging systemic risks.[69] Generally, concerns over the design of platforms’ recommender systems, which aggravated the dissemination of harmful content by prioritizing user-engaging content, were shared among academics and shareholders.[70] The impact assessment[71] refers to Cobbe and Singh, who argue that rather than regulating harmful content, which, in itself, if viewed by a small audience, is unproblematic, regulators should target recommender systems, which ‘can contribute to systemic problems’. Cobbe and Singh note that, by subjecting recommender systems to regulation, regulators can avoid unnecessary infringements of the freedom of expression.[72] Accordingly, the impact assessment notes that there is a lack of understanding on how content is disseminated and supports informing consumers on how these systems work.

Article 29(1) Commission draft DSA obliges very large online platforms to inform consumers of ‘the main parameters used in their recommender systems, as well as any options for the recipients of the service to modify or influence those main parameters that they may have made available’, in their terms and conditions. In addition, very large online platforms shall make available at least one option not based on profiling. This latter obligation should help to limit the dissemination of personalized harmful content and protect users’ private autonomy.[73] However, the Commission could have imposed more obligations, for example by introducing principles of compliance by design. The Commission’s choice to target recommender systems rather than adding these additional obligations can be seen in light of its wider tendency to focus on illegal content rather than expanding consumer rights.[74]

A weakness that becomes apparent, however, is the impact assessment’s lack of attention to the question of how to inform consumers. Although the impact assessment notes the need to make ‘meaningful information’ on recommender systems available, Article 29(1) Commission draft DSA obliges very large online platforms to make information available in their terms and conditions, albeit ‘in a clear, accessible and easily comprehensible manner’. This has, rightly so, led to criticism in literature, as including information in standard terms is not likely to actually inform consumers.[75] After all, it is widely known that consumers typically fail to read standard terms.[76]

Accordingly, other measures such as Article 5(2) P2B, Article 7(4a) Unfair Commercial Practices Directive, and Article 6a(1)(a) Directive 2011/83 on consumer rights include the obligation to include separate, easily available information. Article 3 ELI Model Rules, inspired by both Article 5 P2B Regulation and Articles 5(1) and 6(1) Consumer Rights Directive, also provides that information should be ‘clear and presented in a comprehensible manner, and in machine-readable format. Contract terms must be easily available to platform users at all stages of their relationship with the platform operator’.[77] Although the impact assessment expressly refers to the P2B,[78] Article 29 Commission draft DSA does not contain a similar obligation, nor does it oblige to explain specifically which search results were shown and why.[79] In addition, like other impact assessments in EU consumer law, the impact assessment does not consider behavioural research or legal design in designing information obligations, even though this may significantly enhance the chance that consumers read and comprehend information provided to them.[80]

2.3.  Evidence, strengths and weaknesses in the impact assessment

The impact assessment used an extensive amount of data. It focused on addressing systemic risks through regulating recommender systems, rather than platforms’ terms and conditions. This focus should help prevent violations of freedom of expression, as it shifts the target from individual expressions towards the dissemination of such expressions. Theoretically, the negative effects of this strategy on users whose rights are violated because content is not removed, should be reduced by the limited dissemination of such content.

However, the impact assessment suffers from several serious shortcomings. For one, like other impact assessments,[81] the impact assessment neglected to assess the regulatory choices for regulating standard terms and the impact thereof on the exercise of fundamental rights. Accordingly, Article 12 Commission draft DSA was criticized for its potential detrimental effect on the protection of fundamental rights, and its focus on users who uploaded harmful content rather than the victims.[82] It is not clear whether the Commission draft DSA would have been detrimental, as the requirements in the Unfair Terms Directive do require fair, clear, and predictable terms.

Further, the impact assessment did not consider how consumers should receive meaningful information on recommender systems. Accordingly, Article 29 Commission draft DSA was criticized for its potential ineffectiveness in empowering consumers.[83] This weakness can be contributed to the nature of the impact assessment, which, like other impact assessments, focusses on the problem analysis and the policy options. This focus leaves little room to assess how provisions should be worded – for example, it is beyond the scope of the impact assessment to assess in detail the impact of choices on when or how to inform consumers, or to address changes to the terms and conditions. However, these choices can significantly affect the chance that consumers will actually read and process information.[84]

3.   The European Parliament’s amendments

3.1.  Amendments to Commission draft Article 12 on platforms’ standard terms

Draft Article 12 was subjected to numerous amendments that did however not appear specifically targeted at mitigating the Commission’s impact assessment’s shortcomings listed in this Article. The Committee on Civil Liberties, Justice and Home Affairs did argue that the DSA should target illegal, rather than legal but harmful content, but did not propose different regulatory choices, or leaving harmful content out of the DSA altogether.[85] However, Parliament did address the applicability of fundamental rights to content moderation (Section 3.1.1.), and strengthened criteria for the application and enforcement of standard terms (Section  3.1.2.). The scope and timing of information obligations towards users (Section 3.1.3.) are also relevant for the questions whether possible limitations to users’ fundamental rights are acceptable.

3.1.1.   Amendments on the direct effect of fundamental rights

The justification to proposed amendment 964 refers to the freedom of contract as a fundamental right, referring to the Charter.[86] Proposed amendment 938[87] inserted that providers of intermediary services act ‘in compliance with procedural safeguards and in full respect of the rights and legitimate interests of all parties involved, including the applicable fundamental rights of the recipients of the service as enshrined in the Charter and relevant national law’.[88] The justification thereto clarifies that this amendment seeks to ensure direct applicability of rights in the Charter between platforms and users of their services. It refers to ‘law and jurisprudence’, particularly the Bauer case, where the Court ruled that Article 31(2) Charter, which limits working hours and guarantees rest periods and paid leave, could be invoked in disputes between employees and employers.[89] Other amendments stipulated that standard terms must be in conformity with fundamental rights.[90] While Article 14(4) DSA may aim to ensure that platforms respects users’ fundamental rights, it is still debated whether the Charter can be directly applied in these cases,[91] or affect the interpretation of Article 14 Final DSA. The justifications and the referral to Bauer give the impression that this was Parliament’s intention, as well as recital 47 to the Final DSA, but the amendments were still criticized as insufficiently clear on the question whether platforms should apply fundamental rights law.[92]

3.1.2.   Amendments to apply and enforce standard terms

Whereas the impact assessment only superficially addressed conditions for limiting the freedom of speech, the European Parliament adds these conditions, in line with the ECHR and CJEU case law, human rights reports and recommendations, and surrounding EU legislation. They coincide with stakeholder suggestions and academic criticism of the draft DSA[93] and academic research on the problematic use of standard terms by platforms.[94] The ECHR has repeatedly ruled that a wide discretion for executive bodies to block content, which may result in arbitrarily blocking content that is not illegal, is problematic.[95] This standard allows for some discretion, if exercised in a predictable, consistent manner.[96]

The transparency requirement should be interpreted in line with EU case law on unfair terms, which requires that consumers are able, from the terms of the contract, to assess their future position,[97] as well as the idea, in the P2B Regulation, that plain and intelligible terms ‘give business users a reasonable degree of predictability on the most important aspects of the contractual relationship’.[98] This standard likely satisfies the requirement of foreseeability developed under the ECHR. The standard of fairness should also help to address concerns over platforms’ unfair terms.[99]

These requirements are reflected in various forms. Amendments 82 and 928 add that ‘[p]roviders of intermediary services shall use fair, non-discriminatory and transparent contract terms’,[100] amendment 932 adds ‘straightforward and understandable’,[101] amendment 934 adds the requirements ‘plain, intelligible’,[102] amendment 944 requires that platforms act ‘in a fair, transparent, non-discriminatory’ manner[103] and amendment 939 requires that platforms act in a non-arbitrary manner.[104]

These proposed amendments were partially upheld, reflected in draft amended Article 12(1), (1a), (2), (2d) and (2e).[105] Article 14(3) final DSA requires platforms to ‘act in a diligent, objective and proportionate manner’ in the moderation of content, with respect to the fundamental rights of parties as enshrined in the Charter.

3.1.3.   Amendments to further inform consumers

Amendments targeted both the timing and the form of the obligation to inform users of content moderation policies in an accessible format.[106]

The amendments are in line with Article 3(2) P2B Regulation that provides a reasonable and proportionate notice period of at least 15 days, considered in the 2021 study commissioned by the European Parliament,[107] and the progress report of the Expert Group for the Observatory on the Online Platform Economy.[108] The justification to proposed amendment 956,[109] which would have introduced an obligation to notify media providers of proposed changes to their standard terms, as well as their algorithms or parameters, expressly refers to the P2B Regulation. Proposed amendment 960, which suggested a similar notice period, expressly referred to the Council of Europe Recommendation[110] and the European Parliament’s legislative resolution that also referred to the P2B Regulation and the Unfair Terms Directive.[111] In comparison, Google’s earlier response to the Commission consultation[112] referred to the P2B Regulation, noting that ‘[i]t would be prudent to wait to see its impact before considering any new rules on transparency’. Busch and Mak also noted the need to limit overlap between obligations in Article 29 Commission draft DSA, and corresponding provisions in the P2B, the Unfair Commercial Practices Directive and the Consumer Rights Directive.[113]

Proposed amendment 84,[114] which did not refer to any sources but was similar to these amendments, was partially upheld by Parliament – amended draft Article 12(1b) provided that platforms ‘shall notify expeditiously the recipients of the service of any significant change to the terms and conditions and provide an explanation thereof’. Most of this amendment is visible in Article 14(2) final DSA that obliges platforms to inform users ‘of any significant change to the terms and conditions’. This amendment was welcomed in light of frequent changes to platforms terms and conditions,[115] but the DSA does not clarify what changes are ‘significant’ changes, nor does it determine when and how consumers should be informed of these changes. Possibly, this may mean 15 days, as consideration 10 in the preamble to the DSA and Article 2(4)(e) clarify that the DSA should be ‘without prejudice’ to the P2B Regulation.

The format of information provided to users also improved in line with Parliament’s amendments. Proposed amendment 954 suggested that very large online platforms would be obliged to publish their terms and conditions in all official languages of the EU.[116] Amendment 929 adds that the terms and services should be provided to users of the service in the languages in which the service is offered,[117] as also suggested by EDRI.[118] Accordingly, Article 14(6) final DSA requires that ‘[v]ery large online platforms and very large online search engines within the meaning of Article 33 shall publish their terms and conditions in the official languages of all the Member States in which they offer their services’.

Amendments 924, 929, 931, 935, 936 and 962 proposed the use of summaries in language that is user-friendly, machine readable and concise.[119] The justification to amendment 931 clarifies that requiring machine readability is in conformity with surrounding EU legislation on digital matters.[120] Proposed amendments 935 and 962 obliged platforms to inform users of available remedies and proposed amendment 964 included an information obligation regarding the available dispute mechanisms in Articles 17 and 18 Commission draft DSA.[121] The use of very short summaries that contained information on remedies and independent alternative dispute resolution, was suggested by proposed amendments 925[122] and 936,[123] and recommended by BEUC[124] and AccessNow.[125] The amendments are reflected in Article 14(5) final DSA, which stipulates that platforms shall provide ‘a concise, easily-accessible and machine-readable summary of the terms and conditions, including the available remedies and redress mechanisms, in clear and unambiguous language’.

3.2.  Amendments to Commission draft Article 29 on recommender systems

The draft DSA aimed to address ‘emerging societal risks’ by imposing additional information obligations regarding recommender systems, which met with support and amendments from Parliament. These amendments aimed to widen the scope of obligations, making them more specific, expanding responsibilities conferred on very large online platforms to online platforms (Section 3.2.1.) and proposed ways to inform consumers beyond terms and conditions (Section 3.2.2.). The improvements to the Commission draft DSA further included proposed amendments that sought to address the dissemination of harmful content through recommender systems (Section 3.2.3.), expanded information obligations (Section 3.2.4.) and provided consumers with the possibility to influence recommender systems (Section 3.2.5.). The justifications to (proposed) amendments to Article 29 Commission draft DSA rarely refer to external sources,[126] but align with various authoritative sources and prominent stakeholders, such as recommendations from the EDPS, stakeholders, surrounding EU law and a study commissioned by EU Parliament.

3.2.1.    Amendments widening the scope of obligations regarding recommender systems

Proposed amendments 1687 and 1689 replaced Article 29 Commission draft DSA on recommender systems with draft amended Article 24a, to widen the scope of this provision from very large online platforms to online platforms.[127] The justification to the amendment does not refer to external sources, but is in line with the suggestions of the lead committee’s Rapporteur,[128] and criticism of the draft DSA.[129] The wider applicability of transparency is in line with French law, particularly Article L 117-I which contained a wide definition of platforms, and, in (II)(1) introduced obligations to provide consumers with information that was ‘loyale, claire et transparente’, including information on ranking and referencing content, without making distinctions based on platforms’ size. Article 91(2) Medienstaatsvertrag also had a wider scope than the Commission draft. Article 91(2) Medienstaatsvertrag made an exception for intermediaries averaging less than a million users in Germany in a six month timeframe – this seems a wider exception than Article 16 Commission draft DSA.[130] Article 4 ELI model rules also impose transparency obligations on all ranking systems.[131] The extension of transparency obligations was supported by BEUC.[132] Other stakeholders, such as Verbraucherszentrale Bundesverband, suggested separate rules for recommender systems of online marketplaces,[133] DigitalEurope asked for possibilities for very large online platforms to contest their status,[134] and News Media Europe supported further transparency obligations for very large online platforms.[135] Draft amended Article 24a was adopted by Parliament, and became Article 27 in the final DSA, widening the scope of these obligations beyond platforms facilitating transactions as defined in the Unfair Commercial Practices Directive, the Consumer Rights Directive and the P2B.[136] This, however, did not address possible overlap with the information obligations in these instruments.[137]

3.2.3.   Amendments on how to inform consumers

The impact assessment overlooked research that indicates that consumers – who are users of online platforms, but are by no means the only users – do not read standard terms.[138] The need to inform consumers in more effective ways than standard terms is reflected in surrounding EU law,[139] and national law.[140] A 2021 study recommended providing this information ‘with clearly visible, easily accessible information regarding their options, and adjust their settings in regular intervals’.[141] The EDPS went further and strongly recommended that information on recommender systems be provided separately, easily accessible, clear for average users and concise’.[142] Helberger et al supported this suggestion.[143] Likewise, Amnesty International criticized the use of standard terms and conditions and recommended advising users of detailed parameters, ‘in an easily comprehensible and accessible manner’, separately from terms and conditions.[144]

Proposed amendment 1518 obliged platforms to inform users in draft Article 24a in their terms and conditions, albeit in a ‘clear, accessible and easily comprehensible format’.[145] Proposed amendments 1691 and 1692 additionally introduced the need to inform consumers via standard terms and conditions, ‘on a designated web page that can be directly reached and easily found from the very large online platforms’ online interface’.[146] In comparison, proposed amendment 130 obliged online platforms to set out main parameters and other relevant information in their terms and conditions, but also when content is recommended, ‘in a clear, accessible and easily comprehensible manner’. The timing of this obligation is also relevant, as according to this draft amended Article, users would have been informed dynamically, at a time relevant for them. DotEurope emphasized ‘explainability’ to users, so that consumers would know why they are being shown content, and business would understand rankings, but did not consider possibilities to influence the parameters.[147]

These amendments are largely reflected in draft amended Article 24a(1), which includes that online platforms ‘set out in their terms and conditions and via a designated online resource that can be directly reached and easily found from the online platform’s online interface when content is recommended, in a clear, accessible and easily comprehensible manner’. In addition, draft amended Article 29 included the obligation for very large online platforms to provide ‘an easily accessible functionality on their online interface’ rather than the standard terms suggested in Commission draft Article 29.[148] These draft amendments constituted a significant improvement, but they are not reflected in Article 27(1) Final DSA, which obliges online platforms to inform users via terms and conditions, albeit in ‘plain and intelligible language’.

3.2.3.   Amendments promoting privacy by default and compliance by design

In contrast to the Commission’s tendency to target illegal content rather than expand consumers’ remedies or rights in addition to the GDPR, Parliament proposed more expressly added obligations in line with compliance by design and privacy by default,[149] as well as bans on dark patterns.[150]

Accordingly, proposed amendment 130 targeting recommender systems prohibits platforms from using profiling without users’ ‘freely given, specific, informed and unambiguous consent’, and recommender systems not based on profiling should be the default option – wording which is reminiscent of Article 2(11) GDPR that defines consent as ‘any freely given, specific, informed and unambiguous indication of the data subject’s wishes’, as also supported by Amnesty International.[151] Similarly, proposed amendment 1690 added that online platforms shall ensure that consumers are not profiled by default, unless they ‘genuinely’ opt in, and prohibits using manipulative techniques regarding consumers’ decision-making,[152] in line with the recommendation of the EDPS,[153] Article 19,[154] BEUC,[155] and Access Now.[156] Zuiderveen Borgesius expressed general doubt on the effect of normal opt-ins, as ‘[p]eople tend to click OK to almost any request they see pop up on their screen. This is all the more true when firms make the use of service conditional on the user’s consent’.[157] Fundacja Panyptykon, instead, suggested further transparency on profiling criteria.[158] Despite these amendments, the obligation to provide a recommender system not based on profiling remains limited to very large online platforms, visible in draft amended Article 29.

Proposed amendment 130 in addition suggested that very large online platforms should ensure that recommender systems are not designed in a way that enables misleading consumers. The justification thereto explains that this is in line with privacy by default and compliance by design, to help prevent ‘nudging into the direction of unscientific propaganda, abusive content or conspiracy theories’.[159] The rapporteur of the Committee on Women's Rights and Gender Equality additionally proposed a regular review of algorithms ‘with a view to minimize negative effects on the users’.[160] Further, proposed amendment 1525 added an obligation in proposed Article 24c for platforms to act neutrally, prohibiting them from promoting political parties, opinions, or ideas.[161] This latter amendment was not adopted by Parliament, or reflected in the DSA.

Proposed Article 24a(6) provides that for search results, reliable information (provided by public authorities or scientific research) shall be prioritized. This is not only in line with the recommendation of the IMCO’s rapporteur,[162] and a 2021 study commissioned by Parliament,[163] but also reminiscent of Article I (Purposes) sub vii 2018 Code of Practice on Disinformation which provides that signatories ‘invest in technological means to prioritize relevant, authentic, and accurate and authoritative information where appropriate in search, feeds, or other automatically ranked distribution channels’.[164] The proposed Article could in addition be seen in light of Article 11(2) Charter and the state’s obligation to ensure media diversity,[165] and corresponding recommendations from the High Level Expert Group on Media Freedom and Pluralism to promote digital and media literacy.[166] The idea of diversity has further been encouraged by the Council of Europe that emphasizes that content should be ‘capable of promoting a critical debate and a wider democratic participation of persons belonging to all communities and generations’.[167] Previously, the EDPS had recognized the potential influence of social media to either encourage or discourage voting, as well as other influence on voting.[168] Notwithstanding widespread support, the amendment could also have had adverse effects. Article 19 pointed out that efforts to prioritize some content may adversely affect diversity, as it is frequently minority voices who are less visible as a result.[169]

These amendments were only partially adopted by Parliament and did not amend Article 29 Commission draft DSA. Draft amended Article 13a provided rules for the design of online interfaces for all intermediaries in order to prevent dark patterns.[170] According to the report of the Greens/EFA,[171] this draft Article can be traced to amendment 1014 that aimed to prohibit the use of dark patterns,[172] which was also supported by the Tracking-free Ads Coalition.[173] Article 25 final DSA prohibits providers of online platforms to design their online interfaces in such a way as to deceive, manipulate or otherwise impairs or distorts the decision-making of users, but, in paragraph 2, excludes practices from its scope that fall under the UCPD and the GDPR, rendering the DSA a subsidiarity to these instruments, even though these instruments do not define dark patterns and are ill-suited to address the use of dark patterns.[174] This amendment however coincides with the arguments of Eurocommerce argued the DSA should be complementary to the GDPR and the e-Privacy Directive.[175]

3.2.4.   Amendments to further inform consumers on parameters

Proposed amendment 130 introduced Article 24a(2) which obliged platforms to inform users of parameters used for recommender systems in their standard terms, in line with the GDPR. Proposed Article 24a(3) specified which information to include: the relevant parameters used in the recommender systems, how they are weighed against one another, optimalisation goals, and ‘if applicable, an explanation of the role that the behaviour of the recipients of the service plays in how the relevant system produces its outputs’. The EDPS recommended further clarifications on the use of automated systems, to view used profile(s) to curate content, and the possibility to delete any profiles used to curate content.[176] BEUC supported the EDPS’ recommendations, and, like Fundacja Panoptykon,[177] recommended that all parameters, not just the main ones, be provided.[178] Verbraucherszentrale Bundesverband argued for fair recommender systems, with rankings and reviews that should not be influenced by online marketplaces or sponsored ads.[179] The Bundesverband’s view appears to coincide with the prohibition on misleading commercial practices.[180] Prohibiting ranking based on remuneration would go beyond Article 5(3) P2B that indicates that such ranking is permitted, and the possibility to do so should be communicated to business users. Article 4(2) ELI Model rules on online platforms were inspired by both the UCP and the P2B, as well as the CRD.[181]

The limitations on information obligations imposed by Directive 2016/943 on trade secrets, as suggested by proposed amendments 1695 and 1702[182] were subject to some controversy. On the one hand, Digital Europe and DotEurope underlined the importance of protecting commercially sensitive information and preventing manipulation resulting in undetected harmful content.[183] Limitations resulting from the protection of trade secrets have in addition also been considered in the GDPR,[184] the P2B Regulation,[185] and Article 4(1) ELI Model Rules on Online Platforms.[186] On the other hand, MP Geese argued that the testimony of a Facebook whistleblower should have been a reason not to add these exceptions.[187] The Council of Europe explicitly advised that states should establish transparency obligations regarding ‘the public procurement, use, design and basic processing criteria and methods of algorithmic systems’, and not allow legislation on intellectual property or trade secrets to inhibit such transparency.[188]

These amendments are largely reflected in draft amended Article 24a, which includes elements from the list in Article 24(3) suggested by proposed amendment 130 and the list in proposed amendment 1696.[189] Draft amended Article 24a(2) adds that these requirements ‘shall be without prejudice to rules on protection of trade secrets and intellectual property rights’. Draft amended Article 24a(2)(a) and (b) resemble the wording of Article 5(2) P2B Regulation that obliges online search engines to provide ‘the main parameters, which individually or collectively are most significant in determining ranking and the relative importance of those main parameters’. Less elaborately, Article 27(1) Final DSA provides that platforms shall inform consumers of the main parameters used in their recommender systems, as well as any options for the recipients of the service to modify or influence those main parameters, which, according to Article 27(2) shall ‘at least’ include i) the criteria which are most significant in determining the information suggested to the recipient of the service; and ii) the reasons for the relative importance of those parameters. The amendment regarding trade secrets was not included in the final text of Article 27 DSA.

3.2.5.   Amendments empowering users to influence or modify parameters

Parliament also discussed possibilities for users to influence the main parameters used in recommender systems, beyond existing rules on recommender systems and ranking.[190]

Not all proposed amendments supported this possibility. Thus, proposed Article 24a(2), stipulated that ‘[o]nline platforms shall set out […] the main parameters used in their recommender systems, as well as any options for the recipients of the service to modify or influence those main parameters that they have made available’.[191] The amendment adds that, ‘[w]here several options are available pursuant to paragraph 1’, users should be able to select their preferred option for each of the recommender system, at any given time.[192] The ability of users to modify parameters or select options appears to depend on the availability of such options. It seems less likely that the wording of draft amended provision refers to the availability of parameters as platforms will necessarily make use of these. Likewise, proposed amendment 1707 introduced this possibility to influence or modify parameters as an option for very large online platforms, by stipulating that ‘very large online platforms may provide the recipients of the service with the options to modify or influence the optimisation goals and rank or select the recommendation criteria of relevant systems’.[193] The latter proposed amendment is not accompanied by a justification, and the justification to Article 24a does not refer to external sources.

Proposed amendment 1699 suggested that users be able to service or modify the parameters used by recommender systems of very large online platforms, including at least one option not based on profiling, as defined under Article 4(4) GDPR.[194] This option for users is supported by the rapporteur of the associated LIBE Committee, who favoured ‘giving users control over the algorithms prioritizing the information that is presented to them’ with regard to ‘problematic content’.[195] Proposed amendment 1518 similarly introduced draft Article 24a that included options to select or modify parameters used in recommender systems. The proposed amendment is not accompanied by a justification.[196] Both amendments follow the recommendation in the study for the European Parliament on disinformation and free expression that noted the criticism Commission draft Article 29 had been subjected to because of its phrasing, which did not oblige but permit very large online platforms to make these options available, and recommended this be made mandatory.[197] In contrast, the amendments did not follow arguments from stakeholders for more extensive information obligations.[198]

As suggested by both amendments, draft amended Article 24a(3) lists, under (c), ‘if applicable’, the influence of users’ behaviour. In addition, draft amended Article 29 included the obligation for very large online platforms to provide ‘an easily accessible functionality on their online interface allowing the recipient of the service to select and to modify at any time their preferred option for each of the recommender systems that determines the relative order of information presented to them’.[199] Article 27(1) final DSA, however, provides that online platforms shall set out the main parameters used in recommender systems, ‘as well as any options for recipients of the service to modify or influence those main parameters’.

3.3.  The use of evidence in the European Parliament regarding harmful content in the DSA

The analysis of amendments regarding harmful content in the DSA does not demonstrate that Parliament has a clear answer to the question of whether amendments should expressly or implicitly be supported by evidence, and if so, what amendments, and what evidence. Further, the amendments are not, as such, targeted at mitigating shortcomings in the Commission impact assessment.

In the case of the DSA, Parliament has relied on an extensive amount of sources, obtained independently from the Commission. The justifications to proposed amendments to the draft DSA do not refer to the impact assessment, academic research, or stakeholder positions, but, if they refer to sources, they refer to the Charter and EU (case) law.

Amendments that do not refer to sources are in line with a diverse number of sources that appear to have been considered. As it is not clear how amendments have been negotiated in Parliament, it is not clear how exactly amendments were influenced. Thus, the correlation between these sources and these amendments need not mean that amendments originate from stakeholders. However, some similarities become visible.

Sources that appear to have been considered include not only the Charter and EU (case) law, particularly the P2B and the GDPR, but also reports and recommendations from the EDPB, the Council of Europe, the UN Human Rights Council, national law and the ELI Model Law. For example, the amendments to draft Article 12 DSA clarify the obligations of platforms in the application and enforcement of terms, and considerably improve the format of information to be provided, in line with recommendations from the Council of Europe and the UN Human Rights Council, as well as surrounding EU law. Further, German and French law imposing obligations on platforms and ELI Model law on Online Platforms, as well as stakeholder positions are in line with the widely supported amendments to expand the obligations of very large online platforms regarding recommender systems to online platforms more generally. Other amendments, on the possibility for users to influence recommender’s parameters, align with recommendations from commissioned research that considered criticism of the Commission draft DSA.

Some amendments, for example on the obligation to publish terms and conditions in all languages of the Union, were also defended by prominent stakeholders, such as EDRI, BEUC, and AccessNow. Similarly, the limitation in Article 25(2) DSA is in line with arguments of EuroCommerce. For other amendments, disagreements between stakeholders becomes more visible, for example between AccessNow and academic research on the possibility to genuinely opt in to personalized recommender systems.

Especially sources on human rights appear to be underutilized in the Commission’s impact assessment, considering the superficial assessment of fundamental rights impact in the impact assessment. In the case of the DSA, Parliament’s amendments compensated for this particular shortcoming and improved the Commission draft DSA,[200] notwithstanding remaining doubts on the applicability of the Charter in contractual relationships. However, Parliament did not discuss regulatory choices surrounding standard terms. This may not necessarily be reflected in the quality of Article 14 DSA – Husovec, for example, notes that the DSA ‘constrains without taking away all the power of private companies’,[201] and helps to prevent over-removal by its conditional exemption of liability.[202]

It is difficult to establish a link, let alone a causal link, between evidence and amendments if amendments do not directly refer to sources. Although the use of various sources of evidence is visible in Parliaments’ proposed amendments, a causal link between the use of authoritative sources, such as CJEU case law, and the success of amendments cannot be established in general. In some cases, even a correlation between evidence and the success of amendments seems absent. After all, some amendments coincided with authoritative sources, but not all of those amendments were adopted.[203] In contrast, in some cases, amendments regarding notice periods that referred to the P2B Regulation and Council of Europe Recommendations were apparently not upheld, while a similar amendment that did not refer to external sources was upheld.

In addition, evidence does not always lead to provisions that are predictable and consistent – the amended provisions in line with evidence leave room for doubt. For example, the amendments to inform consumers of recommender systems beyond standard terms did not refer to external sources but were in line with widely held views on standard terms. While they were adopted, they were not reflected in the final DSA. Thus, even if amendments are supported by evidence and in conformity with surrounding law, they may still not survive subsequent interinstitutional negotiations.

Consequently, in the case of the DSA, while evidence may have inspired or indeed improved amendments that subsequently improved the DSA, mentioning evidence does not appear to mean that amendments are of better quality and (or) thereby more convincing to fellow Members of Parliament. The similarity of amendments on some points – such as the relevance of fundamental rights – give the impression that this point also enjoyed wide political support.

Further, the amendments to the DSA demonstrate that the mere availability of evidence does not determine what the outcome of the legislative process will be. Like the impact assessment, Parliament drew inspiration from surrounding EU law such as the P2B, but that does not mean it agreed with the Commission draft DSA. Indeed, Parliament proposed amendments more in line with the P2B that were welcomed in academic debate as well.

4.   General observations on evidence-informed EU lawmaking

Theoretically, the European Parliament could compensate for shortcomings in Commission impact assessments by critically screening impact assessments and evidence therein. However, this is not the focus of the European Parliament. This is confirmed by reports from the European Parliamentary Research Service that indicate that Parliament relies on appraisals of Commission impact assessments and appears to take Commission impact assessments as a starting point.[204]

More generally, it seems unlikely that the European Parliament critically screens impact assessments in other legislative procedures, as the role of Parliament is limited, and it would be inefficient to re-do the Commission’s research. The efficient use of resources may also be a reason for Parliament not to re-investigate points on which there is no clear disagreement, and which appear to be well supported. Some shortcomings may also be easier for Parliament to compensate than others. For example, impact assessments should transparently and systematically reflect stakeholder positions, but one may ask whether Parliament is particularly motivated to challenge the presentation of these views in impact assessments, as stakeholders have sufficient opportunity to have their views heard by Parliament. In addition, while this is not visible in the DSA, Parliament may also compensate if a problem analysis shows manifest errors, provided that Members of Parliament consult widely with stakeholders, either through surveys, public hearings, workshops, and meeting with stakeholders,[205] or otherwise.

Although it is not the focus of the European Parliament to compensate for shortcomings in the impact assessments, its amendments on Articles 14 and 27 DSA compensated some of the shortcomings in therein. Particularly, the amendments improved the chance that fundamental rights of users would be respected by platforms, by directly referring to the Charter, prescribing clearer obligations to inform users of content moderation policies and enforcement thereof, as well as of changes to platforms’ terms and conditions. Still, questions on the access to justice in the sense of Article 6 ECHR and freedom of expression in the sense of Article 10 ECHR remain. In addition, Parliament’s amendments widened the scope of platforms obligations to inform users of the parameters in recommender systems, strengthened compliance by design and privacy by default. Here as well, improvements remain possible, as the possibility to influence or modify parameters left much discretion to platforms.

Parliament’s ability to improve proposed acts also benefits from Parliament’s ability to focus on the wording of provisions, while the impact assessment does not have the room to do so, even though the impact of these changes may be significant. For example, informing consumers of recommender systems exclusively through standard terms is likely not very effective. Yet the use of standard terms to inform consumers of recommender systems was maintained despite well-established problems with standard terms.

It is not clear whether Parliament will more widely consider proposals’ impact on fundamental rights – Parliament did emphasize ensuring compliance with fundamental rights, but cited the DSA as an example where the proposal’s impact on human rights were assessed.[206] On the one hand, the Commission impact assessment could be considered more critically, but on the other hand, even if an impact assessment is not perceived as defective, the European Parliament will apparently consider a proposal in light of the fundamental rights of the Charter, which deserves support. In this respect, it is encouraging that the amount of impact assessments that assess fundamental rights are on the rise - the Regulatory Scrutiny Board notes, in its Annual Report, that around 40% of impact assessment addressed the impact of proposals on fundamental rights.[207] However, impact assessments have a well-documented tendency to emphasize quantifiable impacts over non-quantifiable impacts.[208] Furthermore, the amount of impacts that have to be assessed, increases the risk that the assessment of non-monetary impacts, such as impact on fundamental rights, remains superficial,[209] as confirmed by the FRA.[210] Thus, Parliament seems well-placed to consider the impact of proposals on human rights, and non-economic interests such as protection of weaker parties.[211] The LIBE Committee especially has demonstrated attention for human rights in the EU, for example through its annual reports on this topic.[212] This emphasis may perhaps also be considered in light of the notion of ‘good’ evidence – that is not necessarily neutral but may be linked to political ideas.[213] In comparison, the European Commission does not separately consider respect for fundamental rights in its call for evidence to revise the Better Regulation Guidelines, but rather emphasizes the need for ‘simplicity by design’ and swift, as well as evidence-based EU intervention.[214]

The article demonstrates that Parliament made significant amendments to the DSA, both regarding the scope of platforms’ obligations regarding recommender systems, and their obligations regarding their standard terms and moderation policies. Arguably, the amendments could be seen as ‘substantial’ in the meaning of the IIA. While Parliament did not commission a separate impact assessment for these amendments, these amendments were clearly informed by evidence, as it consulted widely and commissioned a significant amount of research.

5.   Implications for reform

The analysis on the use of evidence by Parliament in the DSA has some implications for the future of EU evidence-informed lawmaking more generally, in light of the European Commission’s call for evidence to revise the Better Regulation Guidelines,[215] and the less recent call to renegotiate the Interinstitutional Agreement on Better Lawmaking.[216] There is potential to strengthen evidence-informed EU legislation, for example by better implementation of the Better Regulation Guidelines within the Commission, and publishing draft Commission impact assessments.[217] However, evidence-informed legislation is not likely to be strengthened by introducing dynamic impact assessments for the European Parliament.[218] The analysis of amendments to the DSA indicates that evidence-informed lawmaking may enhance the quality of amendments, but a relation between evidence and politically successful amendments, or amendments of better legislative quality, for example in terms of consistency, cannot be generally established. Consequently, calls to revise better regulation should not only emphasize generating more evidence throughout the legislative process. Imposing obligations to carry out impact assessments should not shift the attention to collecting even more evidence that can only assess the impact of proposed amendments to a limited extent. The extensive use of other forms of expertise by the European Parliament indicates that more targeted assessments than impact assessments may also provide a basis for informed lawmaking.[219]

Rather, firstly, better regulation should distinguish between the quality and relevance of evidence, as well as subsequent use thereof. The impact assessment and the use of justifications already indicate that not all evidence is of equal value. By its broad definition of evidence, the European Commission does not distinguish between, for example, ECHR case law or surveys. While both sources are relevant, they should be weighed differently. The assessment of compliance with human rights forms an integral part of the development of EU legislation, but not as part of evidence-informed legislation. This is not sufficiently acknowledged in better regulation policy. The Better Regulation Guidelines insufficiently distinguish between the relevance and quality of different sources of evidence.[220] The RSB also does not screen human rights impact assessments in a systematical manner. Including an expert on human rights in the RSB seems a step in the right direction.[221] However, reform should go further. Future revised Guidelines should distinguish more clearly between sources of evidence, also considering the quality of evidence, and the selection of evidence in Commission impact assessments should be more transparent.[222] Moreover, the Better Regulation Guidelines should not frame complying with the Charter of Fundamental Rights, as well as ECHR case law, as a commitment to evidence-informed lawmaking necessary to achieve better results. Doing so risks downplaying the importance of the requirements in this Charter, which should not merely be considered evidence that can be disregarded.

Secondly, strengthening evidence-informed lawmaking may be difficult to reconcile with speeding up the legislative process. This latter aim has been underlined by the European Commission, but the tension with evidence-informed lawmaking is insufficiently acknowledged.[223] It is highly doubtful whether obligations on Parliament in a revised IIA to conduct impact assessments – dynamic or otherwise – would speed up the legislative process within Parliament, because it still takes time and expertise to conduct such assessments.

Thirdly, the extent to which evidence can help prevent political disagreement that leads to delay should not be overestimated.[224] The analysis of the proposed amendments to the DSA instead indicates that collection of evidence does not necessarily overcome political disagreements. While evidence may help highlight political choices, it has not, in case of the DSA, completely prevented stakeholders from advocating amendments in their own interests, though evidence-informed legislation should help perhaps diminish the frequency with which this happens.[225] In the case of the DSA, and other cases, better lawmaking initiatives may also be used as an argument to oppose amendments.[226] The choice between proposed amendments, supported by different sources of evidence, indicates that lawmaking, while informed by evidence, may still be the result of a political choice, likely not only for the DSA, but more widely in the European legislative process. Within the European Parliament, amendments frequently contradict one another, probably because they have been founded on different ideas and sources. For instance, sources on trade secrets may well support arguments to limit information obligations that would oblige platforms to reveal business-sensitive information, while sources on surrounding EU consumer support arguments to further extend information obligations. The debate surrounding evidence in Parliament also demonstrates more broadly the possibility that ‘evidence’ – for example stakeholder’s position papers – may point in different directions, and the mere availability of evidence does not, in itself, determine how draft measures must be amended. This involves political choices as well.[227] The potential of better regulation to overcome politization by rationalizing lawmaking has been questioned before.[228]

Fourthly, imposing binding obligations on the European Parliament to conduct systematic impact assessments are problematic, for example if amendments help to ensure that EU law does not facilitate violations of fundamental rights – even if they are substantial amendments. Likewise, it is problematic to require Parliament to produce evidence to support, for example, amendments to full harmonisation even though the Commission has not supported this regulatory choice with robust, if any, evidence.[229] Last but not least, the European Parliament’s position should not be lessened in comparison with stakeholders and experts who are increasingly involved in the drafting of initiatives in an earlier stage.[230] More practically, it may be difficult to pinpoint the exact impact of the amendments in terms of costs and benefits. The level of detail however does not mean that amendments are insignificant – as demonstrated by the potential impact on developing smart information obligations, for example regarding recommender systems. Instead, the European Parliament appears well-placed to more systematically screen Commission impact assessments, particularly regarding human rights impact, which, in turn, may justify further amendments.[231] Methodical screening could be facilitated by strengthening the working relationship between the European Parliament and the EU Agency for Fundamental Rights.

Fifthly, the delegation of rulemaking power in the DSA are symptomatic for the ‘shift in regulatory power’ towards these actors.[232] The European Parliament, in its amendments to the DSA, has proven itself more critical than the European Commission on this point. Further initiatives that delegate yet more power to already powerful private actors deserve more critical and more systematic focus. The position of political parties within Parliament on this point however differs.[233]

These suggestions improve the evidence available for Parliament, in line with the political interests and advantages Parliament appears to have, without imposing an obligation to assess the general impact of proposed amendments.